University of Wollongong Thesis Collections
University of Wollongong Thesis Collection
University of Wollongong
Year 
Towards optimal treatment planning and
novel dosimetry for cancer patients
receiving intensity modulated radiation
therapy
Nicholas Hardcastle
University of Wollongong
Hardcastle, Nicholas, Towards optimal treatment planning and novel dosimetry for
cancer patients receiving intensity modulated radiation therapy, Doctor of Philosophy thesis,
Centre for Medical Radiation Physics - Faculty of Engineering, University of Wollongong,
2009. http://ro.uow.edu.au/theses/3068
This paper is posted at Research Online.
TOWARDS OPTIMAL TREATMENT
PLANNING AND NOVEL DOSIMETRY FOR
CANCER PATIENTS RECEIVING INTENSITY
MODULATED RADIATION THERAPY
A Thesis Submitted in Fullment of
the Requirements for the Award of the Degree of
Doctor of Philosophy
from
UNIVERSITY OF WOLLONGONG
by
Nicholas Hardcastle
BMedRadPhys
Centre for Medical Radiation Physics, Engineering Physics
Faculty of Engineering
2009
c Copyright 2009
by
Nicholas Hardcastle
ALL RIGHTS RESERVED
CERTIFICATION
I, Nicholas Hardcastle, declare that this thesis, submitted in fullment of the requirements for the award of Doctor of Philosophy, in the Centre for Medical Radiation
Physics, Engineering Physics, Faculty of Engineering, University of Wollongong, is
wholly my own work unless otherwise referenced or acknowledged. The document has
not been submitted for qualications at any other academic institution.
(Signature Required)
Nicholas Hardcastle
4 September 2009
Table of Contents
List of Tables . . . . . . . . .
List of Figures/Illustrations .
ABSTRACT . . . . . . . . .
Acknowledgements . . . . . .
Contribution of Collaborators
Publication List . . . . . . . .
Conferences . . . . . . . . . .
Invited Talks . . . . . . . . .
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
1 Introduction
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
1.1 Aims and Objectives . . . . . . . . . . . . . . . . . . . . . . . . . . . .
1.1.1 Evaluation of advantages or disadvantages of IMRT over 3DCRT
for prostate radiotherapy . . . . . . . . . . . . . . . . . . . . . .
1.1.2 Evaluation of biological optimisation tools for prostate IMRT .
1.1.3 Investigation of Volumetric Modulated Arc Radiotherapy (VMAT)
for prostate cancer . . . . . . . . . . . . . . . . . . . . . . . . .
1.1.4 Optimisation of IMRT plans based on the theoretical 'ideal dose'
1.1.5 Investigation of the dosimetric eect of rectal balloon cavities .
1.1.6 Evaluation of in vivo dosimetry of the rectal wall using rectal
balloons combined with a novel MOSFET dosimeter . . . . . . .
1.1.7 Evaluation of the MOSkin and Gafchromic EBT Film for clinical
surface dose verication . . . . . . . . . . . . . . . . . . . . . .
1.1.8 Measurement of collimator leakage for a linac MLC . . . . . . .
1.2 The Journey . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
1.3 Prostate Cancer . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
1.3.1 Prevalance in Australia . . . . . . . . . . . . . . . . . . . . . . .
1.3.2 Staging and grading . . . . . . . . . . . . . . . . . . . . . . . .
1.3.3 Prostate Cancer Treatment . . . . . . . . . . . . . . . . . . . .
1.4 External beam radiotherapy treatment methods . . . . . . . . . . . . .
1.4.1 Three-dimensional conformal radiotherapy . . . . . . . . . . . .
1.4.2 Intensity Modulated Radiotherapy . . . . . . . . . . . . . . . .
1.5 Photon dose calculation methods . . . . . . . . . . . . . . . . . . . . .
1.5.1 Model based dose calculation algorithms . . . . . . . . . . . . .
1.6 Radiobiological modelling and optimisation . . . . . . . . . . . . . . . .
i
vii
xii
xiii
xvi
xviii
xix
xx
xxi
1
1
1
2
3
3
4
5
5
6
8
8
8
8
9
12
12
13
28
29
32
ii
TABLE OF CONTENTS
1.6.1 Mechanisms of cell killing . . . . . . . . . . . . . . . . . . . . .
1.6.2 Linear Quadratic model . . . . . . . . . . . . . . . . . . . . . .
1.6.3 Biologically Eective Dose and Standard Eective Dose . . . . .
1.6.4 The four Rs of radiobiology . . . . . . . . . . . . . . . . . . . .
1.6.5 BED including tumour repopulation . . . . . . . . . . . . . . .
1.6.6 Hypofractionation . . . . . . . . . . . . . . . . . . . . . . . . . .
1.6.7 Tumour Control Probability and Normal Tissue Complication
Probability . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
1.6.8 Equivalent Uniform Dose . . . . . . . . . . . . . . . . . . . . . .
1.7 Measurement modalities . . . . . . . . . . . . . . . . . . . . . . . . . .
1.7.1 Ionisation chambers . . . . . . . . . . . . . . . . . . . . . . . . .
1.7.2 Radiographic lm . . . . . . . . . . . . . . . . . . . . . . . . . .
1.7.3 Radiochromic lm . . . . . . . . . . . . . . . . . . . . . . . . .
1.7.4 Metal Oxide Semiconductor Field Eect Transistor detectors . .
1.8 Disequilibrium region dosimetry . . . . . . . . . . . . . . . . . . . . . .
33
34
36
37
39
39
41
45
46
47
48
49
50
51
2 Rectal dose reduction with IMRT for prostate cancer radiotherapy 55
2.1 Introduction . . . . . . . . . . . . . . . .
2.2 Method and materials . . . . . . . . . .
2.2.1 3DCRT plan . . . . . . . . . . . .
2.2.2 IMRT plan . . . . . . . . . . . .
2.2.3 Evaluation of results . . . . . . .
2.3 Results . . . . . . . . . . . . . . . . . . .
2.3.1 Dose-volume comparison . . . . .
2.3.2 Biological parameter comparison
2.3.3 Delivery eciency comparison . .
2.4 Discussion . . . . . . . . . . . . . . . . .
2.5 Conclusion . . . . . . . . . . . . . . . . .
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
3 Biological optimisation of prostate IMRT plans
3.1 Introduction . . . . . . . . . . .
3.2 Methods and materials . . . . .
3.2.1 Treatment planning . . .
3.2.2 Plan analysis . . . . . .
3.3 Results . . . . . . . . . . . . . .
3.3.1 Dose-volume histograms
3.3.2 gEUD comparison . . . .
3.3.3 NTCP comparison . . .
3.3.4 Delivery eciency . . . .
3.4 Discussion . . . . . . . . . . . .
3.5 Conclusion . . . . . . . . . . . .
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
55
56
58
58
60
61
61
65
71
71
75
76
76
81
81
82
83
83
85
85
86
86
89
iii
TABLE OF CONTENTS
4 Comparison of prostate IMRT and VMAT biologically optimised
treatment plans
91
4.1
4.2
4.3
4.4
Introduction . . . . . . . . . . .
Methods and materials . . . . .
4.2.1 Treatment planning . . .
Plan analysis . . . . . . . . . .
Results . . . . . . . . . . . . . .
4.4.1 Dose-volume histograms
4.4.2 NTCP comparisons . . .
4.4.3 Delivery eciency . . . .
4.5 Discussion . . . . . . . . . . . .
4.6 Conclusion . . . . . . . . . . . .
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
91
93
93
94
95
95
97
100
100
102
5 Optimisation of prostate IMRT plans based on a theoretical 'goal'
dose
103
5.1 Introduction . . . . . . . . . .
5.2 Method . . . . . . . . . . . .
5.2.1 Contouring . . . . . .
5.2.2 Goal dose distribution
5.2.3 IMRT optimisation . .
5.3 Results . . . . . . . . . . . . .
5.4 Discussion . . . . . . . . . . .
5.5 Conclusion . . . . . . . . . . .
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
6.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . .
6.1.1 Dose escalation and rectal balloons . . . . . . . .
6.1.2 The air cavity eect . . . . . . . . . . . . . . . .
6.1.3 Dose calculation in heterogeneous regions . . . . .
6.1.4 Hypofractionation . . . . . . . . . . . . . . . . . .
6.2 Method . . . . . . . . . . . . . . . . . . . . . . . . . . .
6.2.1 Phantom setup . . . . . . . . . . . . . . . . . . .
6.2.2 Treatment plans . . . . . . . . . . . . . . . . . . .
6.2.3 Single elds . . . . . . . . . . . . . . . . . . . . .
6.2.4 Film calibration . . . . . . . . . . . . . . . . . . .
6.3 Results . . . . . . . . . . . . . . . . . . . . . . . . . . . .
6.3.1 Sagittal geometry . . . . . . . . . . . . . . . . . .
6.4 Discussion . . . . . . . . . . . . . . . . . . . . . . . . . .
6.4.1 Single elds . . . . . . . . . . . . . . . . . . . . .
6.4.2 3DCRT, IMRT and helical tomotherapy deliveries
6.4.3 Clinical signicance . . . . . . . . . . . . . . . . .
6.5 Conclusions . . . . . . . . . . . . . . . . . . . . . . . . .
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
6 Rectal balloon dosimetry in prostate radiotherapy
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
103
104
105
107
108
111
111
113
114
114
114
115
116
116
117
117
120
121
122
122
122
129
129
132
134
135
iv
TABLE OF CONTENTS
7 On the feasibility of in vivo real-time rectal wall dosimetry for prostate
radiotherapy
137
7.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
7.2 Methods and materials . . . . . . . . . . . . . . . . . . . . . . . . . . .
7.2.1 MOSFET measurements . . . . . . . . . . . . . . . . . . . . . .
7.2.2 Radiochromic lm measurements . . . . . . . . . . . . . . . . .
7.3 Results . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
7.3.1 Angular dependence correction method 1: Filtering method . .
7.3.2 Angular dependence correction method 2: Dual MOSkin conguration . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
7.4 Discussion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
7.5 Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
137
138
138
141
141
144
148
152
155
8 Novel surface detectors applied to total scalp irradiation with helical
tomotherapy
156
8.1 Introduction . . . . . . . . . . . .
8.2 Method . . . . . . . . . . . . . .
8.2.1 Treatment plan . . . . . .
8.2.2 Transverse measurements .
8.2.3 Surface measurements . .
8.3 Results and discussion . . . . . .
8.3.1 Transverse measurements .
8.3.2 Surface measurements . .
8.4 Conclusion . . . . . . . . . . . . .
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
156
159
159
160
161
165
165
174
177
9 Multileaf collimator end leaf leakage: Implications for wide-eld IMRT179
9.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . .
9.1.1 MLC leaves and carriages . . . . . . . . . . . . . .
9.1.2 Wide eld IMRT with the Varian Millenium MLC .
9.1.3 Wide eld IMRT in the Pinnacle RTPS . . . . . . .
9.2 Method . . . . . . . . . . . . . . . . . . . . . . . . . . . .
9.2.1 Magnitude of end leaf leakage . . . . . . . . . . . .
9.2.2 IMRT eld . . . . . . . . . . . . . . . . . . . . . . .
9.3 Results and discussion . . . . . . . . . . . . . . . . . . . .
9.3.1 Magnitude of end leaf leakage . . . . . . . . . . . .
9.4 Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . .
10 Summary and future work
10.1
10.2
10.3
10.4
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
179
179
180
182
183
183
185
186
186
196
197
Evaluation of advantages or disadvantages of IMRT over 3DCRT for
prostate radiotherapy . . . . . . . . . . . . . . . . . . . . . . . . . . . . 197
Evaluation of biological optimisation tools for prostate IMRT . . . . . . 198
Investigation of Volumetric Modulated Arc Radiotherapy for prostate
cancer . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 198
Optimisation of prostate IMRT plans based on the theoretical 'ideal dose'199
v
TABLE OF CONTENTS
10.5
10.6
10.7
10.8
10.9
Investigation of the dosimetric eect of rectal balloon cavities . . . . .
Evaluation of in vivo dosimetry of the rectal wall using rectal balloons
combined with a novel MOSFET dosimeter . . . . . . . . . . . . . . . .
Evaluation of the MOSkin and Gafchromic EBT Film for clinical surface
dose verication . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Measurement of collimator leakage for a linac MLC . . . . . . . . . . .
Future work . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
10.9.1 Following on from the current work . . . . . . . . . . . . . . . .
10.9.2 Prostate radiotherapy . . . . . . . . . . . . . . . . . . . . . . .
10.9.3 Target denition . . . . . . . . . . . . . . . . . . . . . . . . . .
10.9.4 In vivo dosimetry . . . . . . . . . . . . . . . . . . . . . . . . . .
10.9.5 Summary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
199
200
202
203
204
204
204
205
212
214
Appendices
216
A Ideal dose script
216
B Monte Carlo simulations
222
A.1 Ideal dose calculation script . . . . . . . . . . . . . . . . . . . . . . . . 216
B.1
B.2
B.3
B.4
Overview of simulations . . . . . . . . . . . . . . . . . . . .
Example BEAMnrc input le . . . . . . . . . . . . . . . . .
Example DOSXYZnrc input le . . . . . . . . . . . . . . . .
Comparison of Monte Carlo simulation with measured data .
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
222
226
233
234
C Statistical analysis
236
References
277
C.1 Student's t-test . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 238
C.2 Wilcoxon rank sum test . . . . . . . . . . . . . . . . . . . . . . . . . . 239
C.3 Spearman's rank correlation test . . . . . . . . . . . . . . . . . . . . . . 240
List of Tables
1.1
1.2
2.1
2.2
2.3
2.4
2.5
2.6
3.1
3.2
3.3
3.4
3.5
4.1
4.2
4.3
4.4
4.5
5.1
5.2
5.3
6.1
6.2
The TNM system for prostate cancer grading . . . . . . . . . . . . . .
MLC properties of the Siemens, Varian and Elekta MLCs . . . . . . . .
IMRT optimisation parameters. ROI = Region Of Interest, DVH =
Dose Volume Histogram, ALAP = As Low As Possible . . . . . . . . .
PTV coverage metrics (averaged over all 16 patients) . . . . . . . . . .
Average rectal percentage volumes receiving 25, 50, 60, 70 and 75Gy . .
V25Gy - V75Gy parameter values for Solid Rectum (SR) and Rectal
Wall (RW) contours for 3DCRT and IMRT plans . . . . . . . . . . . .
Average NTCP values for 3DCRT and IMRT plans with statistical signicance . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Average MU per plan averaged over 16 patients . . . . . . . . . . . . .
Conditions and use of the parameter a . . . . . . . . . . . . . . . . . .
Optimisation parameters used in biological IMRT plans . . . . . . . . .
NTCP calculation parameters . . . . . . . . . . . . . . . . . . . . . . .
Average rectal NTCPs over all 16 patients . . . . . . . . . . . . . . . .
Average MUs . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Optimisation objectives for all IMRT and VMAT plans . . . . . . . . .
NTCP calculation parameters . . . . . . . . . . . . . . . . . . . . . . .
Summary of average DVH parameters over the ten patients . . . . . . .
Summary of average NTCPs for IMRT and VMAT plans . . . . . . . .
Delivery eciency: Average required MUs and delivery time for IMRT
and VMAT plans . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Derivation of optimisation contours . . . . . . . . . . . . . . . . . . . .
Calculated rectal gEUDs from goal DVHs using a=3 . . . . . . . . . .
IMRT optimisation parameters . . . . . . . . . . . . . . . . . . . . . .
IMRT and Helical Tomotherapy optimisation parameters . . . . . . . .
Single eld measurements and RTPS calculations of anterior and posterior rectal wall doses with and without rectal balloon cavity. All errors
are the 95% condence interval of the mean. . . . . . . . . . . . . . . .
vi
9
18
59
65
65
67
67
71
77
82
83
86
86
94
95
97
97
100
106
110
110
120
124
vii
LIST OF TABLES
6.3 Measured and planned cavity wall doses. Percentage dierences are
measured-planned normalized to measured dose. Errors quoted are the
95% condence interval. . . . . . . . . . . . . . . . . . . . . . . . . . .
6.4 Measured and planned rectal wall percentage volumes receiving specied doses. Reported error is the 95% condence interval of the mean
of three measurements. . . . . . . . . . . . . . . . . . . . . . . . . . . .
7.1 Helical tomotherapy optimisation parameters. All doses are in Gy. . . .
7.2 Measurement results for anterior rectal wall measurement . . . . . . . .
7.3 Measured and planned doses at the six locations given in Figure 7.2. . .
8.1 Optimization parameters for helical tomotherapy total scalp treatment
8.2 Example of MOSkin data collection spreadsheet. V is the initial threshold voltage, V is the threshold voltage 30s post-irradiation, and V is
the change in threshold voltage. . . . . . . . . . . . . . . . . . . . . . .
125
129
140
143
144
159
0
164
List of Figures
1.1
1.2
1.3
1.4
1.5
1.6
1.7
2.1
2.2
2.3
2.4
2.5
2.6
An example IMRT eld showing the measured intensity levels using an
Electronic Portal Imaging Device (EPID). . . . . . . . . . . . . . . . . 14
The Varian Millenium 120 leaf MLC (courtesy of http://varian.mediaroom.com/le.php/3
+gold.jpg . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 19
Cell survival curve for typical tumour and late responding normal tissue.
/ =10 was used for the tumour curve and / =3 was used for the
late responding normal tissue curve. . . . . . . . . . . . . . . . . . . . . 35
TCP, NTCP and P+ curves showing the sigmoid shape of the doseresponse curves . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 42
The structure of Gafchromic EBT lm (ISP, 2007) . . . . . . . . . . . 50
Schematic diagram of a MOSFET radiation detector . . . . . . . . . . 51
A 6MV depth dose curve for the rst 1.5cm depth in water showing the
steep dose gradient at the surface. The curve was generated using the
BEAMnrc/DOSXYZnrc Monte Carlo package using a voxel resolution
of 100m in the depth direction . . . . . . . . . . . . . . . . . . . . . . 53
Dose distributions for patients #7 and #11. The left image shows the
3DCRT plan and the right image shows the IMRT plan. The dose scale
ranges from 0-80Gy. . . . . . . . . . . . . . . . . . . . . . . . . . . . . 60
Average cumulative DVHs for (a) PTV and Rectum and (b) Femoral
Heads and Bladder. The individual patient DVHs can be found in
Figure 2.3 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 62
Individual patient PTV and Rectal cumulative DVHs for all patients in
the study . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 63
Average Solid rectal DVH vs rectal wall DVH for a) 3DCRT and b)
IMRT plans . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 66
Rectal NTCPs using (a) model parameters n=1.03, m=0.16 and D50=55.9Gy
(b) model parameters n=0.24, m=0.14 and D50=75.7Gy and c) model
parameters n=0.084, m=0.108 and D50=78.4Gy . . . . . . . . . . . . . 69
Rectal NTCP vs percentage of rectal volume contained by the PTV
for (a) model parameters n=1.03, m=0.16 and D50=55.9Gy (b) model
parameters n=0.24, m=0.14 and D50=75.7Gy and c) model parameters n=0.084, m=0.108 and D50=78.4Gy. Spearman's rank correlation
coecient and p values are presented on each chart . . . . . . . . . . . 70
viii
LIST OF FIGURES
3.1 Behaviour of the gEUD and f(gEUD) functions (a) Example DVHs used
for analysis (b) Change in gEUD as a function of a (c) Optimisation
function value as a function of a and (d) Optimisation function value
as a function of gEUD . . . . . . . . . . . . . . . . . . . . . . . . . . .
3.2 Average cumulative DVHs over all 16 patients for a) PTV and rectum
and b) bladder and femoral heads . . . . . . . . . . . . . . . . . . . . .
3.3 Average calculated gEUDs over all 16 patients for the three values of a
used in planning . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
3.4 Rectal NTCPs for all 16 patients calculated with a) NTCP1 b) NTCP2
and c) NTCP3 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
4.1 Example dose distributions for IMRT (left) and VMAT. Dose scale on
the right is in Gy. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
4.2 PTV and rectal DVHs for all 10 patients . . . . . . . . . . . . . . . . .
4.3 Average cumulative DVHs of a) PTV and rectum and b) bladder and
femoral heads for IMRT and VMAT plans. . . . . . . . . . . . . . . . .
4.4 NTCPs for IMRT and VMAT plans for all 10 patients (a) NTCP1 (b)
NTCP2 and (c) NTCP3 . . . . . . . . . . . . . . . . . . . . . . . . . .
4.5 Total MU for all ten patients . . . . . . . . . . . . . . . . . . . . . . . .
5.1 Contours used for IMRT optimisation. Red = 100% zone, Light Red =
95% zone, Orange = penumbral zone, green = scatter zone and purple
= rectum. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
5.2 Goal dose distribution created in MATLAB . . . . . . . . . . . . . . .
5.3 The 'goal' DVH for all 10 patients compared with the seven eld IMRT
DVHs obtained in Chapter 3 with a=3. The seven eld IMRT plan
obtained by optimising based on the 'goal' DVH is also shown ('planned
goal'). . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
5.4 Resultant IMRT dose distribution . . . . . . . . . . . . . . . . . . . . .
6.1 Phantom setup a) acrylic phantom to hold EZ-EM rectal balloon catheter
b) full phantom setup in prone position c) schematic diagram showing
the location of the sagittal lm (in blue) d) schematic diagram showing
the location of the lm spiral (black lines wrapping around inside of
balloon cavity) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
6.2 Planned dose distributions of the IMRT (left) and helical tomotherapy
plans. The dierences in delivery techniques are seen clearly; IMRT is
delivered using seven beams whereas helical tomotherapy is delivered
using multiple smaller beamlets from the full 360deg . . . . . . . . . .
6.3 Sagittal lm results from (a) single laterally incident beam and (b)
single anterior-posterior beam with and without a cavity. The white
lines show the location of the proles. The arrows show the beam
direction. Horizontal error bars on the plan data show the width of the
planned dose voxels. . . . . . . . . . . . . . . . . . . . . . . . . . . . .
ix
79
84
87
88
92
96
98
99
101
106
108
109
112
119
123
125
x
LIST OF FIGURES
6.4 Sagittal digitised lm images and resultant dose proles for a) 3DCRT
b) IMRT and c) helical tomotherapy (HT) delivery techniques. The
colour bar is in absolute dose in Grays. All measurements were scaled
to represent the dose delivered over the total treatment (28 fractions).
The error bars are the standard error of three measurements. . . . . . .
6.5 Measured and planned rectal wall doses and resultant DVH from spiral
lm geometry. (a) represents the dose to the outermost and innermost
loop of the lm spiral and the planned dose to the lm spiral for the
3DCRT plan (d) represents the resultant rectal wall DVH from the
lm spiral and the planned rectal wall DVH for the 3DCRT plan. (b)
and (e), and (c) and (f) represent the same for the IMRT and helical
tomotherapy plans respectively. . . . . . . . . . . . . . . . . . . . . . .
7.1 The MOSkin detector placed on the RadiaDyne rectal balloon . . . . .
7.2 Location of MOSkin detectors around the rectal balloon cavity for the
second set of measurements . . . . . . . . . . . . . . . . . . . . . . . .
7.3 Anterior rectal wall planned dose compared with measured dose over
the duration of the fraction delivery. Note the dose to the MOSkin is
accrued over the total fraction delivery time. . . . . . . . . . . . . . . .
7.4 (a)MOSkin measured rectal wall doses over time for the six investigated
locations around the rectal wall as given in Figure 7.2 and (b) Temporal
dose accumulation for the six locations . . . . . . . . . . . . . . . . . .
7.5 Relative response for face up and face down MOSkin orientations with
one and two layers of CU on the top edge at (a) 1.5cm and (b) 10cm
depth in solid water. The error bars represent the 95% condence interval of the mean. . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
7.6 (a) The response of the two detectors in the dual MOSkin setup. Error
bars (no end cap for D1) are the 95% CI of the mean (b) The average
response of the two detectors. Error bars are the 95% CI of the mean. .
7.7 (a) The I'mRT phantom setup for dual MOSkin and (b) The normalised
measurement (dual MOSkin / ion chamber) for each incident beam angle. The error bars are the 95% interval of the mean for three measurements. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
7.8 The dual MOSkin measured dose compared with the planned dose for
(a) 3DCRT plan and (b) IMRT plan. The error bars represent the 95%
condence interval of the mean of three measurements. . . . . . . . . .
8.1 Resultant dose distributions and cumulative dose volume histogram for
scalp treatment . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
8.2 (a) 10x10cm and (b)2.5x2.5cm eld depth dose curves with MOSkin,
EBT Film and Attix chamber surface measurements compared with
BEAMnrc and Geant4 (Geant4 data courtesy of Oborn (2008), private
communication) MC simulation data. The depth axis is displayed on a
logarithmic scale to show the detail of the buildup. . . . . . . . . . . .
2
2
127
131
139
142
143
145
147
149
151
153
160
166
xi
LIST OF FIGURES
8.3 Surface dose measurements as a function of incident beam angle for (a)
10x10cm eld and (b) 2.5x2.5cm eld. The ratio of the EBT lm to
the MOSkin measurement changes based on angle and eld size. . . . .
8.4 Transverse lm locations and resultant digitised lm images. The black
dotted line shows the location of the phantom edge. The black lines on
sheets 1 and 2 show the locations of the proles shown in Figures 8.6
and 8.7 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
8.5 Buildup curves for (a) EBT lm and (b) EDR2 lm as a function of
length of lm protruding out of solid water slabs and irradiated edge
on parallel to 6MV photon beam central axis . . . . . . . . . . . . . . .
8.6 (a) Cross plane prole of transverse sheet 1 taken 1cm under peg holes
for EBT lm and plan data. (b) Cross plane prole of transverse sheet
2 taken 2.5cm under peg holes for EBT lm and plan data. Zoomed in
section shows rst 1cm depth in phantom. (c) Posterior-Anterior prole
taken across transverse sheet 2 along the centre of the lm for EBT lm
and plan data. The locations of the proles are shown in Figure 8.4 . .
8.7 The same proles as in 8.6 but with EDR2 data . . . . . . . . . . . . .
8.8 Surface EBT lm locations and measured doses. . . . . . . . . . . . . .
8.9 Comparison of MOSkin measured dose and EBT lm surface dose. . . .
8.10 Sample (every third projection shown) of the incident uence sinogram
for one rotation in the centre (superior-inferior direction) of the PTV.
On each chart the abscissa axis is MLC leaf number and the ordinate
axis is relative planned leaf opening times. The MLC predominantly
blocks the central beamlets of the fan beam and allows beamlets through
that are tangential to the scalp. . . . . . . . . . . . . . . . . . . . . . .
9.1 Schematic showing head and neck IMRT treatment using (a) split coaxial overlapped elds and (b) a single wide eld . . . . . . . . . . . . . .
9.2 Wide eld IMRT as applied with the Pinnacle RTPS. All closed leaf
pairs above the topmost section are positioned at the midpoint of the
topmost leaf opening and all closed leaf pairs below the lowermost
section are positioned at the midpoint of the lowermost leaf opening.
Closed leaf pairs that occur between two openings are positioned at the
average of the midpoints of the two nearest leaf openings . . . . . . . .
9.3 The end leaf leakage for a 6MV photon beam measured at a depth of
1.5cm in solid water using EDR2 lm for (a) 0mm gap width (b) and
3mm gap width . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
9.4 Line proles across the end leaf leakage for a 6MV photon beam measured at a depth of 1.5cm in solid water with EDR2 and EBT lm, and
predicted by Pinnacle for the (a) 0mm (b) 0.6mm and (c) 3mm gap
widths. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
9.5 Comparison of a) the Pinnacle predicted and measured doses for the
end leaf leakage and b) FWHM of end leaf leakage peaks as a function
of width between opposing MLC leaves . . . . . . . . . . . . . . . . .
2
2
167
168
169
171
172
173
174
176
181
183
185
187
188
xii
LIST OF FIGURES
9.6 O-axis end leaf leakage for the a) 0mm gap width b) 0.6mm gap width
and c) 3mm gap width . . . . . . . . . . . . . . . . . . . . . . . . . . .
9.7 The geometry of the Millennium MLC leaf was used to determine the
o-axis distances at which ray-lines from the source would begin to pass
through both leaf tips for the 0mm and 0.6mm leaf gaps. . . . . . . . .
9.8 A wide IMRT eld (a) Radiographic EDR2 lm grey scale map at 10cm
depth in solid water (b) RTPS planar dose maps taken at 10cm depth
in solid water of a wide IMRT eld showing end leaf leakage. The lines
shown represent where line proles were taken. . . . . . . . . . . . . . .
9.9 Proles taken across (a) Line 1 in a low intensity shielded region of the
IMRT eld shown in gure 8 and (b) Line 2 in a high intensity region
of the eld. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
10.1 Dose distributions for (a) 3DCRT (sagittal) (b) IMRT (sagittal) (c)
3DCRT (transverse) and (d) IMRT (transverse) plans for Patient 5
including seminal vesicles . . . . . . . . . . . . . . . . . . . . . . . . . .
10.2 Cumulative DVHs for (a) PTV and rectum and (b) bladder and femoral
heads for Patient 5 . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
10.3 Dose distributions for (a) 3DCRT (sagittal) (b) IMRT (sagittal) (c)
3DCRT (transverse) and (d) IMRT (transverse) plans for Patient 6
including seminal vesicles . . . . . . . . . . . . . . . . . . . . . . . . . .
10.4 Cumulative DVHs for (a) PTV and rectum and (b) bladder and femoral
heads for Patient 6 . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
B.1 Incident electron energy spectrum . . . . . . . . . . . . . . . . . . . . .
B.2 Monte Carlo simulation data (MC) and Ion Chamber (IC) data for a
Varian 21EX linac at 1.5cm, 5cm and 10cm depths (a) X direction prole
and (b) % Depth Dose prole for a 5x5cm eld and (c) X direction
prole and (d) % Depth Dose prole for a 10x10cm eld . . . . . . . .
C.1 The normal probability distribution for a mean of 2 and a standard
deviation of 0.5 shown for the interval of 0 to 4. . . . . . . . . . . . . .
2
2
191
192
193
194
206
207
208
209
224
235
237
Towards Optimal Treatment Planning and Novel
Dosimetry for Cancer Patients Receiving Intensity
Modulated Radiation Therapy
Nicholas Hardcastle
A Thesis for Doctor of Philosophy
Centre for Medical Radiation Physics, Engineering Physics
University of Wollongong
ABSTRACT
Modern radiation oncology is constantly improving and becoming more complex.
Novel dosimetric planning, delivery and dosimetry techniques have allowed for improved plan quality and condence in delivery. This thesis is an investigation into
the impacts of novel radiotherapy planning and delivery techniques and the ecacy of
novel dosimetry methods for modern, complex radiotherapy.
The rst part of the thesis involved investigation into novel treatment planning
optimisation techniques for prostate cancer radiotherapy. Advantages and disadvantages of IMRT for simple prostate radiotherapy in the Australian clinical setting is
investigated, showing small gains compared with high quality conformal radiotherapy.
The use of a radiobiological parameter, specically the generalised Equivalent Uniform
Dose (gEUD) was investigated for prostate IMRT optimisation to reduce rectal dose.
The gEUD metric was found to be a useful optimisation objective that provided rectal
dose reductions over the full dose range. The result of the optimisation was heavily
dependent on the value of a (describing organ architecture), with a lower value of a
resulting in the largest reductions in rectal dose. A commercial Volumetric Modulated
Arc Radiotherapy (VMAT) tool was investigated for prostate radiotherapy. Single arc
VMAT plans were compared to static gantry angle IMRT plans for prostate cancer
cases. It was found that VMAT resulted in equivalent target coverage with reductions
in rectal V25Gy. The VMAT plans required on average 18.6% fewer monitor units
and were theoretically up to 3.75 times faster to delivery compared with static gantry
angle IMRT.
The second part of the thesis looked at using modern radiation detectors for verication of treatment dose in regions of electronic disequilibrium. Rectal balloons lled
with air are used for prostate immobilisation and rectal dose reduction in prostate
photon radiotherapy. This introduces an air cavity into the patient, immediately
adjacent to the target. Radiochromic lm was used to show that two commercial convolution/superposition dose calculation algorithms slightly over-predict the anterior
rectal wall dose and under-predict the posterior rectal wall dose. The feasibility of
a novel MOSFET detector, the MOSkin, coupled to a commercial rectal balloon was
investigated for real time in vivo rectal wall dose verication. In this phantom study,
the MOSkin was shown to be an excellent real time dosimeter, with minimal angular
response and reproducible sensitivity. The MOSkin was then used with radiochromic
lm to verify the dose delivered to the skin during total scalp irradiation with helical
tomotherapy. It was shown that the helical tomotherapy RTPS accurately calculated
the dose to surface voxels and that the dose delivered to the skin is less than the
prescription dose, which suggests a bolus may be required to achieve prescription dose
to the skin. Finally, the dosimetric eect of end leaf leakage was investigated for a
commercial multileaf collimator for wide-eld IMRT. It was shown that end leaf leakage can contribute signicant doses to treatment elds, but provided the eects are
quantied it is reasonable to accept these as the allowance of wide elds avoids complicated dual overlapping eld feathering. The commercial RTPS investigated slightly
under-predicts the magnitude of these end leaf leakage dose contributions.
IMRT, tomotherapy, radiochromic lm, radiobiological IMRT
optimisation, MOSFET detectors
KEYWORDS:
Acknowledgements
Undertaking my PhD has been a very enjoyable experience. I have met many intelligent, friendly, funny people over the last three and a half years who have made the
journey all worthwhile.
I would rstly like to thank my thesis supervisors. Thank you to Prof. Peter
Metcalfe, who has been a fantastic mentor for my clinical research. His easy-going
nature and thirst for knowledge (and coee!) have made it a pleasure and an honour
to work for him. His decades of experience and vast knowledge have provided excellent focus for my work; he is always able to get to the heart of the matter. Thank
you to Prof. Anatoly Rosenfeld. His innovative ideas and detailed knowledge and
experience of dosimetric methods were invaluable during the research. I am also extremely grateful for the logistical and nancial support provided by Prof. Rosenfeld
which have allowed me to travel and expand my professional horizons. I would also
like to thank Prof. Wolfgang Tome for his supervision and support of my visits to
the University of Wisconsin-Madison. Prof. Tome is a very motivating supervisor
who encourages the highest standards of research. I am very grateful for the clinical
knowledge and experience I gained working for Prof. Tome and thoroughly enjoyed
my time in Madison.
I thank Dr. Michael Lerch, Dean Cutajar, Dave Zahra and Peter Ihnat for their
many hours spent working on the MOSFET detector systems for my measurements.
This was very much appreciated. Thank you to Dr. Martin Carolan and Dr. Matthew
xvi
Williams for their time and advice during measurements at ICCC and subsequent
input to writing. Thank you to Dr. Kerwyn Foo and Dr. Andrew Miller at ICCC
for their clinical advice and writing assistance with the planning studies. I would also
like to thank Emilie Soisson, Amar Basavatia and David Westerly for their advice and
direct assistance with my measurements in Madison.
I would like to thank all of my fellow students and the sta at the Centre for
Medical Radiation Physics. It has been some of the best years of my (short) academic
career working with you all. To Amir, Amy, Andy, Brad, Dean, Heidi, Ian, Iwan,
Jeannie, Lucky, Mitra, Scott, Scuba, Tony, and all of the undergrads, I thank you all
for the lunch hours playing poker and dice, doing the quiz and the word puzzle and
generally talking rubbish! I will always look back with fond memories of this period
and wish you all the very best for your future endeavours. I would also like to thank
the sta and students at the University of Wisconsin - Madison. To Amar, Dave,
Dongxu, Ed, Emilie, Eric, Karl, Leah, Noah, thank you for your intelligent discussions
and assistance with my work. Working with you all was a pleasure.
I would like to thank Australian Rotary Health for nancial assistance for my PhD.
To my family - Mum, Dad, Nina, Annie, Granny and Granddad and Lindy and
Russell - thank you all for your constant encouragement, nancial support, food and
wine packages and sympathetic ears during this degree. I am extremely lucky to have
you all in my life.
To my beautiful wife Leah, I'm not sure how I managed to get you but I am so
very thankful that I get to wake up next to you each day. Your love, support and kind
words have made this whole process so much easier.
xvii
Contribution of Collaborators
Professor Peter Metcalfe provided advice on experimental design, data analysis and
writing for all chapters. Professor Anatoly Rosenfeld is the inventor of the MOSkin
dosimeters and provided advice on the use of MOSFET dosimeters, MOSFET experimental design and analysis of MOSFET results. Professor Wolfgang Tome provided
assistance with experimental design for the total scalp irradiation and rectal balloon
projects in addition to advice on analysis and writing for the total scalp irradiation,
rectal balloon and VMAT projects.
Dr. Michael Lerch and Dean Cutajar advised on experimental design for MOSFET
measurements and assisted with assembly of MOSFET dosimeters. David Zahra and
Peter Ihnat produced and modied MOSFET probes with MOSkin detectors and the
MOSFET read out system.
Dr. Martin Carolan provided advice on MOSFET measurements as well as timing
data and writing advice for the VMAT project. Dr. Matthew Williams assisted in
experimental design for the MLC leakage project and advised on the writing for this
project. Abdurrahman Ceylan also provided experimental assistance with the MLC
leakage project. Emilie Soisson and David Westerly assisted with the tomotherapy
measurements. Amar Basavatia assisted with phantom design and construction for
the rectal balloon projects.
Dr. Kerwyn Foo and Dr. Andrew Miller provided advice on experimental design,
analysis and writing Chapters 2-5.
xviii
Publications
Hardcastle N, Metcalfe P, Ceylan A & Williams MJ, Multileaf collimator end leaf leakage: implications for wide-eld IMRT, 2007, Physics in Medicine and Biology, 2007,
52 (21), N493-N504
Hardcastle N, Soisson E, Metcalfe P, Rosenfeld AB & Tome WA, Dosimetric verication of helical tomotherapy for total scalp irradiation, 2008, Medical Physics, 35,
5061-5068
Hardcastle N, Metcalfe PE, Rosenfeld AB & Tome WA, Endo-rectal balloon cavity
dosimetry in a phantom: Performance under IMRT and helical tomotherapy beams,
Radiotherapy and Oncology, 2009, 92, 48-56
Hardcastle N, Davies A, Foo K, Miller A, & Metcalfe PE, Rectal Dose Reduction
with IMRT for Prostate Cancer Radiotherapy, Journal of Medical Imaging and Radiation Oncology (In Submission)
Hardcastle N, Tome WA, Foo K, Miller A, Carolan M & Metcalfe PE, Comparison of
prostate IMRT and VMAT biologically optimised treatment plans, Medical Dosimetry
(In Submission)
xix
Conferences
Hardcastle N, Metcalfe P, Lerch MLF, Tome WA, Rosenfeld AB, Feasibility of in vivo
real-time rectal wall measurements of IMRT and tomotherapy with MOSFET detectors, Accepted abstract, Combined Scientic Meeting, Brisbane, 2009
Hardcastle N, Metcalfe P, Davies A, Miller AA, Foo KY, Comparison of VMAT and
IMRT treatment plans for prostate radiotherapy, Accepted abstract, Combined Scientic Meeting, Brisbane, 2009
Hardcastle N, Metcalfe PE, Rosenfeld AB & Tome WA, Dosimetry with an endorectal balloon, Paper presented at Winter Institute of Medical Physics, Colorado USA,
February 2009
Metcalfe PE, Hardcastle N, Sixteen ways to treat a prostate, Paper presented at Winter Institute of Medical Physics, Colorado USA, February 2009
Hardcastle N, Metcalfe PE, Rosenfeld AB & Tome WA, Rectal Wall Dosimetry in
the Presence of an Endorectal Balloon, In Australasian Physical and Engineering Sciences in Medicine; 2008; pp 423
Hardcastle N, Soisson E, Metcalfe PE, Rosenfeld AB & Tome W, Dosimetric Verication of Helical Tomotherapy for Total Scalp Irradiation, In Australasian Physical
and Engineering Sciences in Medicine; 2008; pp 467
Hardcastle N, Foo KY, Davies A, Miller AA & Metcalfe PE, Biological Optimisation of Prostate IMRT Plans, In Australasian Physical and Engineering Sciences in
Medicine; 2009; pp 37-38
Hardcastle N, Metcalfe P, Ceylan A & Williams MJ, Multileaf collimator end leaf
leakage: Implications for wide-eld IMRT, Paper presented at the 2006 Engineering
and Physical Scientists in Medicine Conference, Noosa, QLD, Australia
Rasmussen K, Schubert L, Westerly D, Hardcastle N, Howard S & Tome WA, "A
method of delivering a low dose fraction using a Tomotherapy unit", Poster presented
at American Association of Physicists in Medicine conference, Texas, USA, 2008
xx
Invited Talks
Hardcastle N, Tome WA, Foo K, Miller A, Carolan M & Metcalfe PE and Rosenfeld AB, In vivo dosimetry of IMRT and Tomotherapy beams using MOSkins placed in
endo-rectal balloons, Presented at Joint Scientic Seminar, CMRP & ICCC: Advanced
Radiobiological Planning in IMRT and Tomotherapy and Advanced Stereotactic Radiotherapy, August 2009
Metcalfe PE & Hardcastle N, Comparison of 3D-CRT, IMRT and VMAT prostate dose
plans using Radiobiological endpoints,Presented at Joint Scientic Seminar, CMRP &
ICCC: Advanced Radiobiological Planning in IMRT and Tomotherapy and Advanced
Stereotactic Radiotherapy, August 2009
Hardcastle N, Jones S, Tome WA, Foo K, Miller A, Carolan M & Metcalfe PE, Biologically Optimised VMAT and IMRT for Prostate Radiotherapy, Presented at Australian
Institute of Radiography TAS Branch Winter Educational Weekend, August 2009
Hardcastle N, Jones S, Tome WA, Foo K, Miller A, Carolan M & Metcalfe PE, Biologically Optimised VMAT and IMRT for Prostate Radiotherapy, Presented at New
Zealand Physics and Engineering in Medicine, August 2009
Hardcastle N, Foo KY, Davies A, Miller AA & Metcalfe PE, Clinical use of normal tissue radiobiology in prostate radiotherapy planning: A sixteen patient sample of
EUD optimised IMRT plans, Presented at Stanford University, Feb 2009
Hardcastle N, Foo KY, Davies A, Miller AA & Metcalfe PE, Clinical use of normal tissue radiobiology in prostate radiotherapy planning: A sixteen patient sample of
EUD optimised IMRT plans, Presented at Medical Physics Seminar Series, University
of Wisconsin-Madison, Feb 2009
xxi
Chapter 1
Introduction
Modern radiotherapy is constantly evolving, becoming more complex as new treatment
planning and delivery methods are developed. This presents a challenge that can only
be met with new planning methods and novel dosimetry techniques.
This thesis is an investigation into cutting edge biological optimisation and novel
in vivo dosimetry methods for Intensity Modulated Radiotherapy (IMRT) and new
delivery techniques - helical tomotherapy and volumetric modulated radiotherapy. The
new plan optimisation techniques were investigated for prostate cancer patients and
are presented in Chapters 2-5. Novel dosimetry techniques were applied to regions of
dosimetric interest in various IMRT situations are presented in Chapters 6-9.
1.1 Aims and Objectives
1.1.1 Evaluation of advantages or disadvantages of IMRT over
3DCRT for prostate radiotherapy
IMRT has resulted in superior prostate radiotherapy plans that reduce organ at risk
(OAR) dose whilst maintaining target coverage. There is some clinical evidence sug1
1.1. Aims and Objectives
2
gesting that reduced rectal toxicity is observed when using IMRT over 3DCRT for
prostate radiotherapy (Zelefsky et al. , 2000, 2001; Kupelian et al. , 2002a,b; Namiki
et al. , 2006; Sanguineti et al. , 2006; Veldeman et al. , 2008). Despite this evidence,
the use of IMRT for prostate cancer is still not standard in Australian clinics.
Research Question:
What are the advantages of IMRT plans over 3DCRT plans for prostate
radiotherapy?
This research question is addressed in Chapter 2 with a 16 patient treatment plan
comparison based on physical dose, radiobiological eect and delivery eciency.
1.1.2 Evaluation of biological optimisation tools for prostate
IMRT
Recent developments have seen the introduction of novel methods for optimising IMRT
plans in commercial radiotherapy treatment planning systems (Choi & Deasy, 2002;
Wu et al. , 2002, 2003; Chapet et al. , 2005; Thomas et al. , 2005; Chvetsov et al. , 2007).
Biological end points are now available for use as optimisation objectives. The ecacy
of biological objectives, specically the generalised Equivalent Uniform Dose (gEUD)
model, was investigated. IMRT optimisation with maximum gEUD objectives for
normal tissues was performed with the impact of model parameter variations examined.
Research Question:
How useful is the gEUD function for optimisation of prostate IMRT plans?
This research question is addressed in Chapter 3 with a 16 patient treatment plan
comparison based on physical dose and radiobiological eect.
1.1. Aims and Objectives
3
1.1.3 Investigation of Volumetric Modulated Arc Radiotherapy (VMAT) for prostate cancer
A new IMRT delivery method is now available from major linac vendors. VMAT
delivery involves the delivery of an IMRT plan using a continuously rotating gantry,
generally delivering the plan in a single rotation of the patient (Yu, 1995; Otto, 2008;
Bzdusek et al. , 2009). This has been shown to result in an increase in delivery
eciency (Afghan et al. , 2008; Palma et al. , 2008a).
Research Question:
Are VMAT plans better than conventional IMRT plans for prostate radiotherapy?
This research question is discussed in Chapter 4 with a 10 patient treatment plan
comparison using physical dose, radiobiological eect and delivery eciency.
1.1.4 Optimisation of IMRT plans based on the theoretical
'ideal dose'
The optimisation of IMRT plans is subject to user determination of how good a plan
can get. One question that needs to be answered, before deciding on the optimal IMRT
plan for treatment, is 'Is one getting the best possible solution for this particular
patient?'. The optimal solution of the IMRT optimisation process will achieve the
desired target dose with normal tissues receiving the lowest possible dose. The optimal
IMRT solution is based on the anatomy of the patient (specically that provided by the
planning CT) and the physical characteristics of the delivery technique. Knowledge
of the optimal dose distribution for a given patient's planning CT would allow the
planner to aim directly for the optimal solution and know when further gains in the
optimisation are no longer achievable.
1.1. Aims and Objectives
4
Research Question:
Can the optimal solution to an IMRT plan be used as a basis for IMRT
planning and how close can one get to the optimal dose distribution using
a commercial RTPS?
This research question is addressed in Chapter 5 with a developed algorithm applied
to a series of prostate cases.
1.1.5 Investigation of the dosimetric eect of rectal balloon
cavities
Rectal balloons are used in many radiotherapy clinics for prostate immobilisation
(Teh et al. , 2001; McGary et al. , 2002; Wachter et al. , 2002; Patel et al. , 2003;
van Lin et al. , 2007). A reduction in prostate motion, and rectal toxicity, has been
observed when air-lled rectal balloons have been used in external beam radiotherapy
for prostate cancer (Patel et al. , 2003; Sanghani et al. , 2004; van Lin et al. , 2005b;
D'Amico et al. , 2006; van Lin et al. , 2007). The eect of the balloon air cavity
on the surrounding dose distribution is a potential concern. The dose distribution in
the presence of a commercially available rectal balloon was measured. The accuracy
of two commercial treatment planning systems in calculating the eect of the balloon
cavity was examined for single eld irradiation, 3D conformal radiotherapy, IMRT and
helical tomotherapy plans.
Research Questions:
What is the dosimetric eect of the rectal balloon air cavity on IMRT and
helical tomotherapy deliveries?
How accurately do convolution/superposition dose calculation algorithms
calculate the dose in the balloon cavity region?
5
1.1. Aims and Objectives
These research questions are answered in Chapter 6 with the use of radiochromic lm
in a rectal balloon phantom.
1.1.6 Evaluation of
in vivo
dosimetry of the rectal wall us-
ing rectal balloons combined with a novel MOSFET
dosimeter
In addition to prostate immobilisation and reduced rectal toxicity, rectal balloons also
provide a means for in vivo dosimetry of the rectal wall. The utility of rectal balloons
coupled with novel Metal Oxide Semiconductor Field Eect Transistor (MOSFET)
radiation detectors as in vivo dosimeters was investigated for conventional prostate
IMRT delivery and helical tomotherapy.
Research Question:
How can an in vivo dosimeter be implemented in a commercial rectal balloon?
A proof of principle dosimetric study looking at MOSFET dosimetry of the rectal wall
is presented in Chapter 7.
1.1.7 Evaluation of the MOSkin and Gafchromic EBT Film
for clinical surface dose verication
Accurate measurement of skin dose in radiotherapy is a challenging task due to the
extremely high dose gradients involved at the patient surface. A novel skin dosimeter
has been developed at the Centre for Medical Radiation Physics (CMRP) - the MOSkin
(Kwan et al. , 2007; Rozenfeld, 2007). The MOSkin is a MOSFET detector with a
reproducible build up layer that provides dose measurement at a water equivalent
depth (WED) of 70m - the ICRP dened depth of the radiosensitive basal layer of
1.1. Aims and Objectives
6
the skin (ICRP, 1991). The MOSkin was characterised in simple radiation elds and
compared to other surface dosimeters. The MOSkin was then used to verify the skin
dose in a complicated radiotherapy treatment; that of a total scalp irradiation using
helical tomotherapy.
Research Questions:
Can an in vivo dosimeter be used on the surface to measure the dose at the
ICRP dened depth of the radiosensitive basal layer of the skin?
Does total scalp irradiation with helical tomotherapy deliver the prescription
dose to the target?
These research questions are addressed in Chapter 8 by the use of dosimetric analysis
using radiochromic lm and MOSFET detectors in conventional radiotherapy linac
and helical tomotherapy beams.
1.1.8 Measurement of collimator leakage for a linac MLC
One vendor's multileaf collimator (MLC) leaves (Varian Millenium 120 leaf MLC)
have rounded ends to ensure a constant penumbral width at all eld locations. This
provides a reduction in radiation path length through the collimator ends when two
leaves are joined without jaw shielding, termed `end leaf leakage' (Boyer & Li, 1997;
LoSasso et al. , 1998; Arneld et al. , 2000b). Jaw shielding is not viable when the
radiation eld is larger than 14.5cm, which occurs in head and neck IMRT and pelvic
node treatments. Leakage through the ends of the collimator can lead to dose 'hot
spots' in the radiation eld. End leaf leakage was characterised and measured for
simple eld arrangements and for a wide eld IMRT treatment.
Research Questions:
What is the magnitude of end leaf leakage through the Varian Millenium
MLC?
1.1. Aims and Objectives
7
Can wide-eld IMRT be used safely without splitting the eld i.e. is the
end leaf leakage maintained at an acceptable low level?
These research questions are discussed in Chapter 9 with a dosimetric analysis using
radiographic and radiochromic lm.
1.2. The Journey
8
1.2 The Journey
The research for this thesis was carried out at the Centre for Medical Radiation Physics
(CMRP) at the University of Wollongong, the Department of Medical Physics at
the University of Wisconsin-Madison and Illawarra Cancer Care Centre (ICCC) at
Wollongong Hospital. The dosimetric studies were carried out in two visits to the
University of Wisconsin-Madison with their helical tomotherapy and rectal balloon
systems. The planning studies were carried out at the CMRP and ICCC as well as at
the University of Wisconsin-Madison.
1.3 Prostate Cancer
1.3.1 Prevalance in Australia
The number of new cases of prostate cancer in Australia was projected to be 18,784 in
2009, making prostate cancer the most prevalent of all cancers, followed by colorectal
(14,405) and breast (13,805) cancer (AACR, 2008). Prostate cancer is expected to
aect one in three males before the age of 75 and one in two males before the age
of 85. It is projected that prostate cancer will kill 3,283 people in 2009; the fourth
deadliest cancer in Australia.
1.3.2 Staging and grading
There are three main systems used for prostate cancer grading and staging. Prostate
cancer staging can be done using the Tumour - Node - Metastasis (TNM) System
(Gospodarowicz et al. , 2004). The TNM System is described in Table 1.1.
The Prostate Specic Antigen (PSA) test is a blood test that determines the extent
of the tumour. PSA is a protein produced by the prostate. Levels above 4ng/L have
1.3. Prostate Cancer
9
Table 1.1: The TNM system for prostate cancer grading
Grade
Description
T1
Tumour is small and cannot be felt by the doctor
T2
Tumour can be felt, but is still conned to the prostate
T3
Tumour can be felt, but may have invaded the seminal vesicles
T4 Tumour has invaded other organs/tissues in the surrounding pelvic region
N1-3
Tumour has invaded the lymph nodes in the pelvis
M1
Tumour cells present in bone and other distant organs
shown to be an indicator of prostate cancer. Of men with PSA levels of 4-10ng/L,
25% have cancer and of men with PSA level above 10ng/L, 60% have cancer (Stewart
et al. , 2003). However, it has been found that PSA level association with cancer may
vary amongst races (Assessment, 1997). This has lead to the conclusion that PSA
level alone is not sucient for determining the presence of prostate cancer (Institute,
2007).
Prostate cancer grading comes from a biopsy of the tumour tissue, and determines
how abnormal and how aggressive the tumour is. Grading is commonly performed
using the Gleason Score (ranging from 2-10), with faster growing tumours given a
higher score. TNM Staging, PSA level and the Gleason Score can be combined to
assess the risk of recurrence and risk of the cancer spreading to other organs. These
three factors are also taken into account when determining the appropriate treatment.
1.3.3 Prostate Cancer Treatment
1.3.3.1 External Beam Radiotherapy
External Beam Radiotherapy (EBRT) involves the use of a radiation beam incident
externally on the patient to irradiate the prostate. The radiation beam is generated
most commonly by a linear accelerator (linac), but can also be delivered using Co-60
sources. For the remainder of this thesis EBRT is discussed in terms of linac delivered
1.3. Prostate Cancer
10
photon radiation.
The total radiation dose prescribed by the oncologist is commonly split up into
a number of fractions, with each fraction delivered separately, commonly with a day
between fractions. The delivery of the total dose in multiple fractions, rather than one
single fraction, is due to the radiobiological property of normal tissue and tumours
whereby small single doses damage tumours more than normal tissue and large single
doses damage normal tissue more than tumours. This is explained in detail in Section
1.6.2.
Modern EBRT consists of three steps - patient imaging, treatment planning and
radiation delivery. Patient imaging is the acquisition of a planning computed tomography (CT) scan of the target anatomy. The oncologist denes the region to be scanned
and the desired resolution and contrast. A planning CT scan is taken to obtain a
volumetric image of the patient's anatomy. The CT data is then transferred to a
Radiotherapy Treatment Planning System (RTPS).
The RTPS includes software that acts as a virtual treatment and allows for target
denition and calculation of the expected treatment dose. Traditionally treatment
planning systems allowed for 'virtual simulation' of the treatment in two dimensional
image space using simulated beams eye views created with a linac simulator. The
virtual simulation tools have since evolved into modern three dimensional treatment
planning. Modern treatment planning involves the delineation of the target and normal
tissue anatomy on the planning CT and then creating a treatment plan using a RTPS.
The oncologist outlines the target and any organs at risk on the CT data set using
the RTPS software. A series of treatment beams are then created in the software and
the expected dose is calculated. Once an acceptable plan is created, that achieves
sucient target dose while minimising normal tissue dose, the plan data is transferred
to a Record and Verify (RV) system. The plan data consists of the location of the
1.3. Prostate Cancer
11
patient with reference to the linac frame of reference in the treatment room and the
parameters that dene the motion and delivery characteristics of the linac.
For delivery of each radiotherapy fraction, the patient is placed on the treatment
couch. The patient is then aligned to the treatment position using lasers, portal images
or volumetric images (such as in-room CT scanners) so that the target location is in
the same location as in the treatment plan. Once the patient is set up, the RV system
veries the radiation is delivered as per the treatment plan.
1.3.3.2 Brachytherapy
Brachytherapy is the delivery of radiation to a tumour volume using radiation sources
placed inside the target. Brachytherapy can be split into two main types - low dose
rate (LDR) and high dose rate (HDR). Brachytherapy is an invasive procedure but
requires less total time for delivery. Brachytherapy delivers a highly conformal dose
to the target. Brachytherapy follows the same work ow as EBRT, that is, patient
imaging, treatment planning and radiation delivery.
LDR Brachytherapy is achieved using lower activity seeds that contain radioisotopes. These are surgically implanted into the target at locations dened by the treatment plan. Most LDR treatments use 60-120 Iodine-125 sources, however Paladium103 are also used (Williamson et al. , 2005). The seeds are implanted permanently
and continuously irradiate the target at a low dose rate over a number of half lives
of the isotope. After a sucient number of half lives of the isotope have passed, the
seeds are no longer delivering treatment dose.
HDR Brachytherapy involves the use of an afterloader. A series of catheters are
placed surgically into the target volume, according to the treatment plan. An afterloader is then used to direct a high activity source into each catheter, in series. Most
HDR prostate Brachytherapy uses a stepping Iridium-192 source with a nominal activity of 10Ci. The treatment plan determines at what depths and how long the source
1.4. External beam radiotherapy treatment methods
12
dwells in each catheter, so as to irradiate uniformly the target to the required dose.
Commonly for prostate HDR Brachytherapy, the total dose is delivered in one to three
fractions, up to eight hours apart. The catheters are left in between fraction delivery.
1.3.3.3 Combined Brachytherapy and EBRT
Brachytherapy and EBRT are commonly combined using an EBRT treatment to a
given dose and then an HDR treatment used as a boost to the prostate. An example
treatment schedule is conformal radiotherapy given to the prostate to a dose of 50Gy
followed by three fraction of HDR Brachytherapy at 9Gy/fraction.
1.3.3.4 1.2.3.4. Other Treatment Approaches
Low grade prostate cancer can be quite slow to grow. As a result, for many patients
a 'watch and wait' approach may be taken. This could be used in a case where
the possible side eects of treatment would outweigh any benets. Other treatments
include cyrotherapy (cooling of the tumour), hormone therapy and more recently high
intensity focused ultrasound therapy (Kennedy, 2005).
1.4 External beam radiotherapy treatment methods
1.4.1 Three-dimensional conformal radiotherapy
Three-dimensional Conformal Radiotherapy (3DCRT) evolved from four eld box
treatments due to the introduction of 3D treatment planning. Target delineation
using 3D data sets has allowed the projection of the shape of targets to be transferred
to the shape of the beam. 3DCRT is then the process whereby multiple radiation
1.4. External beam radiotherapy treatment methods
13
beams are collimated to irregular shapes such that a highly conformal dose is delivered to the target and surrounding normal tissues are shielded. The weights of the
beams i.e. their individual contribution to the total dose is then optimised to achieve
sucient target coverage and normal tissue shielding. Wedges are used to account for
non-uniformities in the radiation path length through the patient to achieve a more
uniform dose distribution (Webb, 1993). Compensators can also be used to achieve
uniform target dose distribution to take into account missing tissues.
1.4.2 Intensity Modulated Radiotherapy
Intensity Modulated Radiotherapy (IMRT) is the creation of non-uniform intensity
distributions delivered to the tumour from multiple beam directions. This is achieved
through the use of either physical compensators that dierentially attenuate the beam
depending on the location in the eld, or through collimator devices such as multileaf collimators (MLCs). IMRT diers from 3DCRT in that 3DCRT involves shaped
elds that have a uniform cross-eld dose distribution (sometimes including a wedge),
whereas IMRT involves shaped elds that have purposely modulated cross-eld dose
distributions. The creation of an IMRT plan can either be forward or inverse planned.
Forward planning is an iterative, manual customisation of the beam angles and intensity. Inverse planning is an iterative process that utilises optimisation algorithms
that optimise the intensity proles of each eld to achieve a desired dose distribution (Webb, 1989). Figure 1.1 shows the measured dose of an IMRT eld using an
Electronic Portal Imaging Device (EPID). The non-uniform intensity of the eld is
clear.
1.4. External beam radiotherapy treatment methods
14
Figure 1.1: An example IMRT eld showing the measured intensity levels using an
Electronic Portal Imaging Device (EPID).
1.4.2.1 IMRT Planning Process
IMRT plans are created by designated IMRT RTPS software that contains inverse
planning tools. The patient CT data is imported and target organs and normal tissues
are delineated. The user then denes a number of beam energies and gantry angles, as
would be done for a 3DCRT plan. After the beams have been selected the user then
sets the optimisation objectives to describe the desired target and OAR doses. Optimisation objectives take the form of physical dose and biological end points. Common
physical dose objectives include maximum dose, minimum dose, mean dose and maximum and minimum dose volume objectives, where maximum or minimum volumes
receiving given doses are set. Biological objectives include maximum, minimum and
target generalised Equivalent Uniform Dose (gEUD) (Niemierko, 1997) (for the Pinnacle RTPS) and mathematical models describing the probability of tumour control or
normal tissue toxicity (for the CMS Monaco RTPS) (Alber & Nusslin, 1999). The user
then selects the optimisation algorithm (described in Section 1.4.2.2), the maximum
number of segments or control points and the method of conversion to a deliverable
1.4. External beam radiotherapy treatment methods
15
uence map. The optimisation is then run whereby the algorithm searches for the
optimal uence prole to meet the optimisation objectives. IMRT optimisation is an
iterative process in that the user can constantly update objectives to meet desired
dose objectives.
1.4.2.2 Optimisation Algorithms
There are a number of optimisation algorithms available however two main categories
exist. In the rst type, an ideal uence map is found for each beam which is then
converted to a uence map that can be delivered by the linac using the desired compensator limitations. The second type, direct aperture optimisation (DAO), is where
the solution is made up of a machine deliverable uence from the start. That is,
the optimiser solution space contains only uence maps that are within the linac and
compensator limitations.
IMRT optimisation is based on an objective function that returns an objective
value based on deviation of the current plan from dened optimisation parameters.
The higher the objective value the worse the plan, the goal is to minimise the objective
function. Optimisation parameters describe the desired dose distribution and include
physical or biological parameters, as described above. Optimisation is an iterative
process whereby each eld in the plan is split into a number of smaller segments. The
weighting on each segment is then iteratively adjusted to achieve a dose distribution
that satises the objectives. An objective function is used to calculate the deviations
from the desired (objective) dose distribution and the current iteration's dose distribution. A quadratic function is often used to calculated the weighted sum of the squared
dierence between the actual and goal dose distribution. If the desired dose in a voxel
j is given by dgoal
j , then the objective function for N voxels can be given as (Metcalfe
et al. , 2007):
16
1.4. External beam radiotherapy treatment methods
F
=
N
X
j =1
dgoal
j
sj dj
2
(1.1)
To minimise the objective function, the result after each iteration must be an
improvement on the previous iteration. To achieve this, various techniques are used
to drive down the objective value. These can be deterministic or stochastic.
An example of a deterministic method is the gradient minimisation method (Spirou
& Chui, 1998). This requires a concave or convex objective function whereby the gradient of the objective function is minimised (concave) or a gradient descent algorithm
is used (convex). For a concave objective function, successive iterations reduce the
gradient of the change in the objective function until the rst derivative of the function is equal to zero i.e. the objective function cannot be reduced any further. The
changes are made by modifying individual beamlet weights such that the optimisation
function value is reduced. The change in a beamlet weight for beamlet i is given as
(Metcalfe et al. , 2007):
wj =
dF d2 F
=
dwi dwi2
!
(1.2)
Which, using equation 1.1, becomes:
wj =
PN
dgoal
Dji
j
PN
2
j =1 sj Dji
j =1 sj
dj
(1.3)
The term is a damping term to ensure convergence to a solution. Weights cannot
be < 0, therefore any weights that do violate this are set to equal zero. The weight
change for a given beamlet becomes a weighted average of the dose dierences for all
voxels inuenced by the beamlet, where the weights are dependent on the optimisation
objective weights of each voxel.
A stochastic method is simulated annealing (Webb, 1989). Simulated annealing
1.4. External beam radiotherapy treatment methods
17
applies Boltzman statistical mechanics principles of atoms in a solid to the optimisation
problem. Advantages of stochastic methods are that it avoids solutions in a `local
minimum' where a gradient reduction method returns a solution in a local minimum
rather than the global minimum (Jeraj & Keall, 1999). Due to their random nature
however, stochastic methods require many more iterations.
A more recent development in IMRT optimisation algorithms is Direct Aperture
Optimisation (DAO) (Shepard et al. , 2002). DAO combines intensity modulation
and leaf sequencing into one step. The incident uence can be described by the leaf
positions and the beamlet weighting. The leaf parameters are incorporated into the
optimisation function so that the solution obtained is directly deliverable and doesn't
need converting to a deliverable uence. DAO allows the user to limit the minimum
MU per segment and minimum aperture/segment size, potentially decreasing delivery
time.
DAO is implemented into the Pinnacle RTPS as Direct Machine Parameter Optimisation (DMPO). IMRT optimisation with DMPO involves setting the eld energies
and orientations, the optimisation objectives and segmentation properties such as minimum segment size and MU. The rst few iterations then nd a set of control points
that satisfy machine and objective constraints. The remainder of the iterations optimise the MLC leaf positions and segment weights. Therefore, the uence is deliverable
at all times during the optimisation. During the iterations, a pencil beam dose calculation is used to calculate the update dose. At the start, the end and at set iteration
number during the optimisation, a full collapsed cone convolution dose calculation with
a deliverable uence is performed. This increases the accuracy of the dose calculation
during the optimisation and leads to a solution that better meets the optimisation
objectives (Hardemark et al. , 2003).
1.4. External beam radiotherapy treatment methods
18
Table 1.2: MLC properties of the Siemens, Varian and Elekta MLCs
Vendor
Mount
Number of Leaf width at
Leaf end
location
leaves
isocentre
shape
Siemens Replaces lower jaw
160
0.5cm
Double focused
a
Varian
Below Jaws
52, 80 or 120 1cm & 0.5cm
Rounded
Elekta
Above jaws
80
1cm
Rounded
a 1cm for the outside 40 leaves and 0.5cm for the central 20 leaves
1.4.2.3 Delivery techniques
Delivery of IMRT with a physical compensator
is where the modulated eld is described as a two dimensional (2D) matrix of intensity
values. The intensity values are then converted into a 2D map of thicknesses through
a given compensator material. The compensator is then manufactured for each eld
using the exported dimensions from the RTPS. During delivery, a separate compensator must be manually placed in the block tray of the linac head prior to delivery of
each eld.
1.4.2.3.1 Physical compensator
A multileaf collimator (MLC) consists of a bank
of collimator `leaves' that move in and out of the radiation eld, collimating the beam
to a given shape. An example of a commercial MLC is shown in Figure 1.2. Each leaf
can travel beyond the centre of the radiation eld and have the ability to `interleave',
producing complicated shapes. MLCs allow for eld shaping between and during each
radiation eld delivery, allowing highly conformal doses to be delivered to the target.
A summary of the various vendor's MLCs is presented in Table 1.2.
MLCs were originally designed for dening open elds and as a result, when they are
used to create complicated modulated elds for IMRT, certain dosimetric properties
become apparent:
Tongue and groove eect: On either side of a leaf there is a tongue or a groove,
1.4.2.3.2 Multileaf collimator
1.4. External beam radiotherapy treatment methods
19
Please see print copy for image.
Figure 1.2:
The Varian Millenium 120 leaf MLC (courtesy of
http://varian.mediaroom.com/le.php/301/MLC+-+gold.jpg
1.4. External beam radiotherapy treatment methods
20
which are in place to lock leaves together to minimise inter-leaf leakage. In IMRT
segments, when the tongue side of a leaf denes the outside of a segment, additional
shielding occurs due to the protrusion of the tongue into the open segment. This can
overlap with other tongue protrusions in other segments resulting in an underdose
region (Siochi, 1999).
Matchline eect: For MLCs with rounded leaf tips, the radiation path length
through the leaf decreases towards the distal end of the leaf. This widens the eld
penumbra. When radiation through the leaf ends overlaps in multiple segments of
a step and shoot IMRT eld, high dose lines become visible. These are known as
matchlines (Cadman et al. , 2002; Tangboonduangjit et al. , 2004).
End leaf leakage: When two opposing rounded leaf tips are joined in the radiation
eld, leakage occurs. This is generally not a problem when closed opposing leaves are
shielded by jaws. There are however some instances, such as wide eld IMRT, where
end leaf leakage occurs in the treatment eld with no jaw shielding. This is discussed
in Chapter 9.
The dosimetric properties of MLCs must be taken into account by the RTPS so that
underdosing or overdosing does not occur. A number of commercial RTPSs have quite
elegant leaf models with some studies published investigating the optimal parameter
settings (Cadman et al. , 2005; Williams & Metcalfe, 2006).
Leaf sequencing algorithms are designed
to convert an ideal uence distribution into a uence distribution that is physically
deliverable by a commercial linac MLC. The uence distribution of a modulated eld
consists of a 2D map of intensity values. There are various algorithms under two
categories - dynamic (Spirou & Chui, 1994; Stein et al. , 1994; Svensson et al. , 1994)
and step and shoot (Ma et al. , 1998; Xia & Verhey, 1998). Dynamic delivery uses
dynamic MLCs, that is, the MLC leaves are moving constantly while the beam is on.
1.4.2.3.3 Leaf sequencing algorithms
1.4. External beam radiotherapy treatment methods
21
Step and shoot delivery is where the modulated eld is made up of multiple static
elds (segments) with the beam switched o, while the MLC leaves move between
segment positions. Leaf sequencing algorithms in their basic form take the optimal
uence of each eld in the form of an intensity matrix. A linear combination of binary
matrices, equivalent to the intensity matrix, is then calculated (Siochi, 1999).
Sliding window is the delivery of a modulated
eld using a changing aperture shape that moves across the eld, resulting in a modulated dose distribution. Sliding window IMRT can be delivered using the step and
shoot technique (BORTFELD et al. , 1994) or dynamically (CONVERY & ROSENBLOOM, 1992).
1.4.2.3.3.1 Sliding window
K-means clustering is a leaf sequencing algorithm that groups intensity levels in the ideal uence distribution into `K-clusters'.
It then nds the optimal distribution of intensity levels minimising dierences between
the ideal and deliverable uence (Wu et al. , 2001). K-means clustering is used by
the Pinnacle RTPS for Varian and Elekta MLCs. There are two steps in the process
- grouping of the intensity levels and optimisation to minimise discrepancies between
the ideal and deliverable uences.
After a uence map has been obtained from the IMRT optimisation algorithm, all
non-zero intensity levels are grouped into a minimum number K-clusters such that the
maximum dierence between two intensity levels in a given cluster is less than a given
error tolerance. An optimisation algorithm is then initiated to calculate a given set
of intensity levels, forming an approximated uence, by averaging each cluster. The
discrepancies between the original ideal uence map are then calculated and minimised.
The last step, eld decomposition (segmentation), takes the approximated K-clusters
of intensity levels and converts them to an MLC deliverable uence based on physical
1.4.2.3.3.2 K-means clustering
1.4. External beam radiotherapy treatment methods
22
limitations of the leaves. There are many algorithms to perform the segmentation
(Boyer & Yu, 1999; Wu et al. , 2001).
IMFAST was developed as a beam sequencing algorithm
designed to minimise the delivery time of static modulated elds (Siochi, 1999). The
delivery time for a given modulated eld is optimised based on beam on time, leaf travel
and record and verify (R&V) overheads. The leaf travel distance is reduced whilst
still taking into account leaf collision and tongue and groove eect. Two methods are
used to reduce leaf travel distance - extraction and rod pushing. Extraction reduces
the number of segments by extracting out `matched' (same shape) segments of the
intensity matrix and replacing multiple unique segment shapes with fewer matched
segments. Rod pushing nds an optimal sequence whereby the leaves all move in the
one direction to create the intensity distribution. The leaf sequencing for a complicated
eld is achieved through a combination of extraction and rod pushing such that the
delivery time of a complicated intensity map is minimised.
1.4.2.3.3.3 IMFAST
Tomotherapy is the delivery of IMRT using a slice fan
beam that rotates continuously around the patient. Tomotherapy can be delivered serially with two slices (Peacock NOMOS) or helically (TomoTherapy Inc.) with one slice.
All tomotherapy measurements and discussion in this Thesis concern helical tomotherapy,
delivered with the TomoTherapy Hi-Art (Highly Adaptive Radiotherapy) Helical tomotherapy is delivered with a continuously rotating 6MV photon fan beam. The fan
beam is collimated by jaws in the superior-inferior direction to allow fan widths of
1, 2.5 and 5cm. In the left-right direction the fan beam is collimated by a 64 leaf
binary MLC. The MLC leaves are pneumatically driven such that they can be opened
or closed in 20ms. The leaves are binary and cover the full eld width, meaning only
one bank of leaves is required. Each MLC leaf projects to a width of 0.625cm at 85cm
1.4.2.3.4 Tomotherapy
1.4. External beam radiotherapy treatment methods
23
from the source. Maximum eld width in the left-right direction is 40cm at 85cm from
the source. As the helical tomotherapy system delivers IMRT only, no attening lter
is present in the beam. As a result, the open eld prole is peaked in the middle.
An inherent advantage of the TomoTherapy beam delivery system is its ability to
obtain megavoltage CT (MVCT) scans. The waveguide is detuned to provide a 3.5MV
fan beam and an array of xenon gas eld ionisation chambers, mounted opposite to the
fan beam, collect transmission data to acquire a volumetric image set. The acquisition
of an MVCT prior to delivery of each treatment fraction is incorporated into the
TomoTherapy workow. A second inherant advantage of the TomoTherapy delivery
system is the highly modulated dose distributions that can be achieved. This is due
to the rotational IMRT delivery and the binary 64 leaf MLC. These provide many
degrees of freedom for intensity modulation.
The TomoTherapy Hi-ART system consists of an RTPS, an R&V system and
the beam delivery hardware (linac). The RTPS process involves importation of the
planning CT data, contouring of the target and OARs and inverse planning. Inverse
planning in TomoTherapy is similar to conventional IMRT however a number of extra
parameters are required as a result of the helical, modulated fan beam delivery.
Collimator width: This is the thickness of the fan beam. Selection of this value
is chosen as a compromise between level of modulation required and delivery eciency.
A larger collimator width allows a larger volume to be treated per rotation, making
it faster, but limits the amount of modulation that can be performed per rotation.
Values of 1, 2.5 and 5cm are available.
Pitch: The pitch is the couch movement per rotation in units of the eld width.
The pitch also allows the user to control the level of modulation and eciency. A
smaller pitch allows for a greater level of modulation, but does not necessarily mean
a slower delivery time. This is because a smaller pitch leads to a greater number
1.4. External beam radiotherapy treatment methods
24
of rotations which means that a greater overlap between rotations is observed. The
required dose per rotation decreases and the gantry rotation speed can increase.
The choice of pitch can introduce a dosimetric artefact - the helical tomotherapy
thread eect (Kissick et al. , 2005). This is the appearance of dose 'ripples', that occur
due to beam junctioning. This eect is unique to helical tomotherapy. The thread
eect is minimised by using a pitch value equal to p=0.86/n (n is an integer) (Kissick
et al. , 2005).
Modulation factor: The modulation factor set by the user actually represents
the maximum modulation factor used by the optimisation algorithm. The modulation
factor is the ratio of the maximum leaf opening time to the mean leaf opening time for
all MLC leaves used in the treatment. This therefore represents the level of modulation
the optimiser uses. A small modulation factor limits the amount of modulation that
can be achieved and as a result decreases treatment times.
A more recent technique
of delivering modulated dose distributions is Volumetric Modulated Arc Radiotherapy
(VMAT). VMAT is the delivery of a modulated dose distribution using a continuously
rotating gantry with continuous radiation output. VMAT is the latest iteration of
an idea originally proposed by Yu (1995), who suggested Intensity Modulated Arc
Therapy (IMAT) as an alternative to tomotherapy. IMAT provided modulation via
dynamic MLC motion during the gantry rotation. VMAT diers from IMAT in that
VMAT can provide both dose rate and gantry speed modulation. Additional variables
include arc length, avoidance sectors, collimator rotation, couch rotation and the number of arcs. The recent VMAT work was initiated by an article by Otto (2008) and
subsequent commercial implementation by Varian of Otto's algorithm in their RapidArc solution. The term VMAT was used by Otto but has since been trademarked
by Elekta as their modulated arc radiotherapy solution. During this thesis however,
1.4.2.3.5 Volumetric Modulated Arc Radiotherapy
1.4. External beam radiotherapy treatment methods
25
VMAT will be used as a generic term describing any recent modulated arc radiotherapy
technique.
There has been a large push for VMAT research and commercial development since
the paper by Otto, as well as much discussion as to the dierences between VMAT
and static gantry angle IMRT (S-IMRT) and tomotherapy. This is based on the potential decreases in delivery time with VMAT. VMAT is delivering dose during the
gantry rotation where as S-IMRT is only delivering dose from static gantry angles. In
addition, VMAT is delivering the dose volumetrically, rather than in in slices. This
leads to potential decreases in time required to deliver a treatment fraction when compared to S-IMRT and tomotherapy. Recent discussions in the literature have covered
actual delivery time advantages of VMAT over S-IMRT and helical tomotherapy. An
article by Ling et al. (2008) was the subject of much discussion in Int. J. Radiat.
Oncol. Biol. Phys regarding claimed delivery eciency advantages of VMAT (specifically RapidArc) over helical tomotherapy; VMAT was said to be 5-15 times faster
than helical tomotherapy. This was refuted by Mehta et al. (2009), with examples
suggesting that these advantages were over-estimated. Actual delivery eciency will
need to be compared clinically with S-IMRT and tomotherapy for large numbers of
patients to obtain real data for eciency comparisons. At this stage it appears that
although VMAT does have some eciency gains over S-IMRT and tomotherapy, these
gains may have been overstated by some vendors and related publications.
A second point of recent discussion on VMAT has been the level of modulation
that can be achieved with VMAT delivery. Specically, this has been due to some
vendor's claims that one rotation around the patient is all that is required (since
updated to two arcs for complex targets). Bortfeld & Webb (2009) discussed this
point in a technical note in Phys. Med. Biol. where they compared modulation
achieved with S-IMRT, tomotherapy and VMAT with the ideal modulation for the
1.4. External beam radiotherapy treatment methods
26
original target with avoidance structure presented by Brahme (1982). It was shown
that tomotherapy achieved the greatest level of modulation and that single arc VMAT
and S-IMRT had to make compromises in their delivery. This note must be taken in
the context that some delivery parameters and variables were not taken into account
(Otto, 2009). Regardless of the assumptions by Bortfeld & Webb (2009), it has been
shown that some complex geometries require more than one arc to deliver the desired
modulation (Guckenberger, 2009).
A more recent article by Webb & McQuaid (2009) provides some initial mathematical framework of VMAT delivery. At each gantry angle, no modulation occurs.
It is only when apertures at successive gantry angles overlap that modulation of the
dose distribution can occur. This comes from the assumption, presented by Webb
& McQuaid (2009) that if the beamlets from successive gantry angle apertures are
assumed to be parallel rather than diverging, a modulate intensity prole exists. This
is analagous to S-IMRT, where individual segments overlap to provide a modulated
intensity prole at each static gantry angle. It was shown by Brahme (1982) that to
achieve sucient coverage of a donut shaped target with no uence directed towards
the hole, high intensity was required just past the hole and rotation was required.
That is to say, both rotation and modulation was required. This is not achievable
with single arc VMAT, but may be achievable to a limited degree with multiple overlapping arcs. It has been shown that increasing the number of elds improves the dose
distribution (Jones, 1999). However, simply increasing the number of elds cannot
achieve the level of modulation required for complex, concave targets, as described
above. Modulation from each discrete beam angle is still necessary.
Commercial and research
VMAT optimisation algorithms all rely on the approximation of a continuously rotating gantry with successive static gantry angles. This 'small-angle' approximation
1.4.2.3.5.1 VMAT planning and optimisation
1.4. External beam radiotherapy treatment methods
27
has been shown to result in low dosimetric error, from the point of view of the target,
but can have larger errors at the periphery of the patient (Webb & McQuaid, 2009).
This thesis uses Pinnacle's SmartArc VMAT tool, which was developed by RaySearch
Laboratories and is also used in the Oncentra MasterPlan RTPS. The resultant VMAT
plans can be delivered on both Varian and Elekta linacs. The SmartArc algorithm has
been described and evaluated in a recent publication (Bzdusek et al. , 2009). Other
VMAT algorithms, such as that used for Varian RapidArc, vary in the method used to
achieve the nal modulated arc however all are based on starting with an initial coarse
angular spacing which is resampled to achieve a ne gantry angle spacing (required for
accurated dose calculation). The SmartArc optimisation process with the SmartArc
algorithm is as follows:
The initial arc parameters (such as arc length, delivery time, number of arcs)
are set by the user
The arc length is split into a nite number of elds, spaced equally around the
arc with a separation of 24
Intensity modulation is performed on each eld, resulting in intensity maps
spaced every 24
Intensity maps are converted into 2-4 leaf and jaw segments per map(using sliding
window). The segments satisfy static machine constraints
The segments are distributed around the arc length. This is done by taking the
two segments with the highest number of leaf pairs and repositioning them one
third of the initial angle spacing (24 / 3 = 8) to the left and right of the initial
angle
Segments are then created and inserted at angles evenly between the existing
segments to match the user-selected nal angle spacing. These segments are
created by linearly interpolating between the existing segments. This results in
the desired gantry angle spacing, with a segment (or control point at each angle).
For example, for an arc length of 360 and a nal gantry angle spacing of 4,
91 control points are created and placed every 4 around the arc. There are 91
control points as the arc is not the full 360, rather 359.9. Therefore an extra
control point is required to describe the control point for the gantry angle 359.9
1.5. Photon dose calculation methods
28
The machine parameters for the MLC and jaw segments are then optimised using
a gradient based algorithm. The optimisation takes into account gantry speed,
dose rate, total arc delivery time and maximum leaf travel speed
The jaw positions are set. For machines with static jaws, the jaw positions are
set to the maximum segment size. For machines with dynamic jaws, the jaw
positions are set for each segment
For optimisation of two arcs, the initial intensity maps are converted into 4 segments, all of which are kept. The four segments are then distributed two to an arc.
The distribution is based on the position of the segment relative to the centre of mass
of the target, such that each arc aims to deliver dose to dierent halves of the target.
During the optimisation, the algorithm employs a modied pencil beam dose calculation method - the Singular Value Decomposition method (Bortfeld et al. , 1993).
This decreases the dose calculation time during the iterations. Full collapsed cone
convolution calculations are performed during the optimisation (if selected by the
user) and at the end of the optimisation iterations (McNutt, 2002). Segment weight
optimisation is also performed on the nal segments.
1.5 Photon dose calculation methods
The accuracy of a treatment plan IMRT optimisation depends heavily on the accuracy of the dose calculation algorithm. There are two categories of dose calculation correction based and model based (Mackie et al. 1996). Correction based algorithms
start with calculating the dose analytically in a simple situation such as a water cube.
Various corrections are then applied to the dose distribution based on patient shape
and size, heterogeneities in the patient density, beam characteristics and beam modiers. E-depth (Bentley & Milan, 1971) and E-TAR (Sontag & Cunningham, 1978) are
examples of correction based algorithms. Model based algorithms model the incident
photon uence from the treatment head and then use analytical models to calculate
1.5. Photon dose calculation methods
29
the dose in the specic patient's anatomical data set. In general, model based algorithms are more accurate than correction based algorithms in low density regions such
as the lungs. As model based algorithms constitute the majority of commercial treatment planning system algorithms, including those used in this Thesis, a description of
model based algorithms only is presented.
1.5.1 Model based dose calculation algorithms
As stated above, model based dose calculation involves modelling the exact treatment
head dimensions and characteristics, as well as any beam modifying devices. An accurate model of the incident photon uence is then derived, which is then projected onto
the patient's volumetric anatomical data set (CT data set) and the dose is calculated.
Examples of model based dose calculation algorithms include convolution/superposition (C/S) algorithms (Boyer & Mok, 1985; Mackie et al. , 1985a), pencil beam convolution algorithms (Mohan et al. , 1986), the Anisotropic Analytical Algorithm (AAA)
(Ulmer & Harder, 1995, 1996) and Monte Carlo techniques. The model based dose
calculation algorithms used in this thesis are the convolution/superposition algorithm
and Monte Carlo.
1.5.1.1 Convolution/superposition algorithm
The C/S algorithm is used by the Pinnacle and TomoTherapy RTPSs. The c/s algorithm is recognised as one of the most accurate model based dose calculation algorithms
(Arneld et al. , 2000a; Carrasco et al. , 2004; Jones & Das, 2005; Vanderstraeten et al.
, 2006; Fogliata et al. , 2007). The C/S method was initially proposed by Mackie et al.
(1985a) and was developed in parallel by Mackie et al. (1985a) and Ahnesjo et al.
(1987). The C/S algorithm contains three steps - calculation of the incident uence,
projection of the uence through the patient's CT data to obtain the TERMA volume
1.5. Photon dose calculation methods
30
and convolution of the TERMA volume with dose deposition kernels to obtain the
absorbed dose (Ahnesjo & Aspradakis, 1999). The steps are detailed below.
Calculation of the incident uence: The rst step in modelling of the treatment head is to obtain the incident photon uence. Each element of the treatment
head is modelled in the RTPS including the target and source size, beam energy spectrum, attening lter, primary collimator, jaws and multileaf collimators. Any beam
modifying devices such as compensators and wedges are modelled. The result of this
step is a 2D uence map representing the incident beam.
Calculation of the TERMA volume: Once the incident uence has been obtained, the Total Energy Released per unit MAss (TERMA) volume is calculated. The
patient's anatomy is represented by a 3D matrix of CT numbers. Each CT number
corresponds to a given density. Each voxel in the CT data set is converted to density.
The density volume is then converted into a series of photon attenuation tables based
on the density in each voxel and the energy spectrum of the beam. The incident uence
is then ray-traced through the volume, and at each step (each voxel encountered by a
ray), the energy loss is calculated using the attenuation coecient tables. Softening
of the beam spectrum due to dierential attenuation is also taken into account. The
end result is a 3D matrix of TERMA values, that is, a 3D matrix of the total energy
released per unit mass due to the incident uence.
Convolution of the TERMA volume and dose deposition kernels: The
nal step is the convolution/superposition of the TERMA volume and dose deposition
kernels. A dose deposition kernel can be thought of as a 3D table of absorbed dose, due
to a primary interaction, at vectorial displacements relative to a primary interaction
site (Mackie et al. , 1988). The kernels can also be thought of as an analytical function
describing the absorbed dose in each voxel of the patient due to a primary interaction
in a single voxel (Ahnesjo, 1989). The absorbed dose at any point away from a primary
1.5. Photon dose calculation methods
31
interaction site is a result of scattered radiation and secondary particles put in motion
by the primary interaction.
Kernels can be created using Monte Carlo methods; a number of monoenergetic
photons are forced to interact in a given voxel and the scattered dose, due to secondary
particles, is recorded in the surrounding voxels (Mackie et al. , 1988). This is repeated
for a given number of energy bins covering the whole photon energy spectrum of the
beam. The result is a series of monoenergetic kernels. The rst comprehensive set of
Monte Carlo generated dose deposition kernels was created by Mackie et al. (1988).
For a full volume dose calculation in each voxel in the 3D data set, the TERMA
value is convolved with the dose deposition kernels to obtain the dose in each voxel of
the patient's anatomy. In order to speed up this time consuming process, the Pinnacle
and TomoTherapy RTPS employ a collapsed cone convolution (Ahnesjo, 1989; Ahnesjo
& Aspradakis, 1999). Collapsed cone convolution is where all of the energy released
into coaxial cones of equal solid angle is `collapsed' into the central axis of the cone.
That is, all of the energy within the cone is transported, attenuated and deposited
into volume elements along the central axis of the cone (Ahnesjo, 1989).
1.5.1.2 Monte Carlo dose calculation
Monte Carlo simulation is a computational method used to model stochastic phenomena, such as radiation transport. Monte Carlo simulation generates random numbers,
which are then used to sample random variables based on probability distributions
governing radiation transport. Monte Carlo simulation of radiation transport uses
analytical physics models of particle interactions to model the energy deposited in a
given medium, hence the absorbed dose deposition. Monte Carlo simulation of radiation transport requires modelling of Compton scattering, the photoelectric eect, pair
production and inelastic and multiple scattering of particles.
Monte Carlo simulation involves simulating individual particle tracks from origin
1.6. Radiobiological modelling and optimisation
32
until they have lost all of their energy, or have a lower energy than set thresholds. For
a given interaction, random numbers are generated. These are then used to sample
interaction probability distributions, particle interaction cross-sections and energy deposition characteristics. Example calculations for a photon track include the distance
to the next interaction (based on energy of photon and medium density), the type of
interaction (sampled randomly from interaction probabilities), new angle and energy
of post-interaction photon (sampled randomly from cross-section tables) and tracking
of any secondary particles created or set in motion during the photon track.
Common Monte Carlo codes include EGSnrcMP (Electron Gamma Shower) /
BEAMnrcMP (Rogers, 1984; Rogers et al. , 1995), GEANT4 (GEometry ANd Tracking) (Agostinelli, 2003), PENELOPE (Baro et al. , 1995) and MCNP(X) (Monte Carlo
N-Particle) (Briesmeister, 1986). For radiotherapy applications EGS/BEAMnrc is the
most common as it has in built geometry designed for linac head modelling. The
EGS/BEAMnrc code only takes into account electrons and photons. The advantages
of Monte Carlo dose calculation are the accuracy of the calculation, the high resolution obtainable and the additional information such as energy spectra, uence, angular
distributions of particles and dose components. The disadvantage of Monte Carlo is
the long time required for calculations which has hindered its uptake in commercial
RTPSs. More recently, so called 'macro' Monte Carlo methods have been developed
that appear to be fast enough for routine treatment planning. One such method is
VMC++ (Kawrakow & Fippel, 2000). Macro Monte Carlo methods are based purely
on lookup tables, rather than analytical generation of radiation interactions.
1.6 Radiobiological modelling and optimisation
The study of the eect of ionising radiation on human cells is termed Radiobiology.
Radiobiological models are designed to predict the eect of ionising radiation on tu-
1.6. Radiobiological modelling and optimisation
33
mour cells and normal tissue cells. Radiobiology mechanisms are very complex, but
there are a number of models based on cell survival data that can predict, to an extent,
the eect of ionising radiation on tumours and normal tissue. The disadvantages of
these models are that they have a very tenuous biological basis, but are derived to t
measured data.
The purpose of radiotherapy is to maximise damage to the tumour clonogens (a
tumour cell from which a tumour can regenerate) whilst minimising damage to normal
tissue cells; there must be a balance between tumour control and side eects. As local
tumour control increases, reduction of side eects must be pursued, particularly when
radiotherapy patients are surviving for longer periods post-therapy (Bentzen, 2006).
It is with radiobiological models that clinicians can tailor dose distributions according
to tumour control and normal tissue toxicity. Understanding of radiobiology, and the
response of cells to ionising radiation, is constantly changing and being updated. The
radiobiological models used in this thesis are derived from 'classical' radiobiological
models; therefore the following discussion only covers the 'classical' models.
1.6.1 Mechanisms of cell killing
Radiotherapy is based on the principal of killing tumour clonogens whilst minimising
complications of surrounding normal tissues. Traditionally, ionising radiation kills or
damages cells via deposition of energy directly in the DNA strands of cells. DNA
strands are killed either directly or indirectly. Direct cell kill or damage usually occurs
when electrons cause double (DSBs) and single strand breaks (SSBs) respectively in
the DNA. Direct cell kill due to DSBs is represented by the initial shoulder region
in a cell survival curve (Section 1.6.2). Although there are mechanisms for repair of
DSBs, not all DSBs are repaired or are misrepaired, leading to cell death. Indirect cell
kill is a result of free radicals, the most important of which is OH-, produced by the
34
1.6. Radiobiological modelling and optimisation
ionisation of a water molecule.
The traditional view of cell kill as being purely due to direct interaction with
DNA strands has been questioned recently (Prise et al. , 2005). Phenomenon such as
low dose hypersensitivity (Joiner et al. , 2000) and the bystander eect (Nagasawa &
Little, 1992; Morgan, 2003; Prise & O'Sullivan, 2009) have suggested that intracellular
and intercellular signalling pathways play an important role in the response of cells to
ionising radiation.
1.6.2 Linear Quadratic model
The most common modern radiobiological model is the Linear Quadratic (LQ) model
(Thames, 1985). The LQ model is derived from cell survival curves for irradiated cells.
The LQ model allows for cell proliferation and repair to be easily incorporated. Cell
survival curves consist of survival fraction versus absorbed dose, as shown in gure
1.3. The LQ calculates the surviving cell fraction S as a function of dose D using both
a linear term and a quadratic term:
S = e(
(d
d2
+
))
(1.4)
The linear term has the coecient , and describes the initial linear section of the
cell survival curve. The linear section of the cell survival curve is related to cell kill
via single radiation events. The second section of the cell survival curve is described
using a quadratic term with the coecient . The quadratic section of the cell survival
curve is related to cell kill through multiple hits.
The coecients and , having the units Gy and Gy respectively, are characteristic of the type of tissue/organ or tumour in question. There is wide variation
in the values of and for given tissues, however the ratio / is commonly 10Gy
for tumours and 3Gy for late responding normal tissue (see section 1.6.6 for more
1
2
1.6. Radiobiological modelling and optimisation
35
Figure 1.3: Cell survival curve for typical tumour and late responding normal tissue.
/ =10 was used for the tumour curve and / =3 was used for the late responding
normal tissue curve.
discussion). Figure 1.3 also shows the reason for fractionation in radiotherapy. At
the high dose end of the scale, the survival fraction for tumour cells is greater than
that for normal tissue. At around 13Gy the curves cross and the survival fraction for
tumour cells is less than that for normal tissue. This shows that to achieve greater
tumour control for less normal tissue damage, the radiation dose must be delivered in
multiple smaller fractions. The common fraction size is 2Gy/fraction, which means
that the survival curve for 0-2Gy is the result of each fraction delivery, resulting in
greater tumour cell kill than normal tissue cell kill. The LQ model for cell survival for
fractionated radiotherapy regimes is easily modied to take into account fractionation.
If one considers the total dose D as the product of the number of fractions n and the
dose per fraction d, then the cell survival S is given as the product of the cell survival
for each individual fraction:
36
1.6. Radiobiological modelling and optimisation
S = e(
S = e(
(d
d2
+
)) n
(1.5)
(1.6)
nd(+d))
1.6.3 Biologically Eective Dose and Standard Eective Dose
Biological Eective Dose (BED) is a method of taking into account radiobiological
properties of tumours and normal tissues when describing dose. The BED can be
calculated for normal tissues and for tumours, and allows for calculation of 'biologically
equivalent' fractionation regimes. That is, calculation of fractionation schedules that
have dierent dose per fraction and number of fractions but the same biological eect.
The BED for a given fractionation regime can be thought of as the equivalent dose
delivered in innitely small fractions. BED is derived by equating the cell survival for
a given fractionation regime with that for innitesimally small fraction size (i.e. full
repair, an abstract quantity) (nd is total dose) (Fowler, 1989):
nd ( + d) = BED ( + 0)
d
BED = nd +
=
!
(1.7)
(1.8)
The Standard Eective Dose (SED) allows for calculation of the BED relative to
a standard fractionation size of 2Gy. The SED is derived similar to BED but instead
of equating the survival fraction with innitesimally small fraction size, the survival
fraction is equated with a 2Gy fraction size:
nd ( + d) = SED ( + 2)
(1.9)
1.6. Radiobiological modelling and optimisation
SED =
37
BED
(1.10)
1 + =
The SED is useful for calculating the eective dose delivered by various fractionation regimes relative to the standard fraction size of 2Gy.
BED and SED calculations become useful when calculating the biological eect of
non-standard fractionation regimes. The BED and SED calculations allow for calculation of new fractionation regimes that either have the same tumour biological eect
for less normal tissue toxicity or greater tumour biological eect for equivalent normal
tissue toxicity. This is most commonly employed in changes in fractionation schedules,
either hypofractionation or hyperfractionation. A description of hypofractionation and
example calculations are given in Section 1.6.6.
2
1.6.4 The four Rs of radiobiology
In order to provide a comprehensive review of the linear quadratic model, the four Rs
of radiobiology, as summarised by Withers (1975), must be discussed. The four Rs
are Reoxygenation, Redistribution, Repair and Repopulation. These four phenomena
contribute to the benets of fractionation, in combination with the shape of the cell
survival curves. The \intrinsic cellular radiosensitivity" of dierent tissues has also
been suggested by Steel et al. (1989) as being a fth 'r' of radiobiology (Zips, 2009).
Reoxygenation: Well oxygenated cells are more sensitive to low Linear Energy
Transfer (LET) radiation than hypoxic cells, requiring approximately 1/3 the dose to
achieve equal cell kill (Thames & Hendry, 1987). As the distance from vasculature
increases, cells become more hypoxic and consequently are harder to kill (Hall 1988).
With fractionation in radiotherapy, the well oxygenated cells are killed leaving blood,
hence oxygen supply, for the hypoxic cells to reoxygenate and increase in sensitivity
(Withers, 1985).
1.6. Radiobiological modelling and optimisation
38
The radiosensitivity of a cell depends on which phase of the cell
cycle is in. The cell cycles through four phases - G1, S, G2 and M (Howard & Pelc,
1953). The G2/M and late G1/early S are the most radiosensitive phases (Metcalfe
et al. , 2007). A given population of tumour cells contains cells in a heterogeneous
range of phases, which changes over the course of a radiotherapy treatment. Extending
the delivery of radiation through fractionation increases the probability that all tumour
cells will be in a radiosensitive phase at one stage of the treatment course.
Repair: Fractionation allows repair of normal cells between fractions. It is suggested that at least six hours between fraction deliveries should be maintained to limit
normal tissue damage (Saunders et al. , 1991).
Repopulation: Cell repopulation is important for both tumours and early responding normal tissue and depends on the cell type. For early responding normal
tissue, cell repopulation increases with time after the rst irradiation, meaning at
later stages of a prolonged fractionation regime, larger doses are required to obtain
the same biological eect (Metcalfe et al. , 2007). Increasing the duration of the treatment course is thus benecial to early responding normal tissue. This is not the case
for late responding normal tissue.
For tumour cells, accelerated repopulation can also occur. After the start of a
course of radiotherapy, the division rate of tumour cells increases. The time from the
start of the course to increase in repopulation is referred to as the 'time for kicko',
Tk . As a result, prolonged treatment time can be detrimental to tumour control.
The rate of repopulation is described by the potential doubling time, Tp. This is the
time required to double the number of tumour cells assuming no cell loss. The ideal
fractionation regime takes into account the values of Tk , Tp and the repopulation
capacity of the normal tissues at risk for a given treatment site.
Recent studies have inferred the presence of tumour 'stem cells' in the tumour cell
Redistribution:
39
1.6. Radiobiological modelling and optimisation
population. These are referred to as tumour cell clonogens. The control of such cells
is a necessity for tumour control if a long term cure is desired (Milas & Hittelman,
2009).
1.6.5 BED including tumour repopulation
The BED can be calculated for a given fractionation regime taking into account tumour
repopulation by modifying the LQ model. The LQ cell survival equation, including
fractionation (Equation 1.6) can be modied to include an exponential cell population
term with a coecient for a total treatment time T:
S = e(
(1.11)
nd(+d)+T )
The term T can be thought of as the loss in cell kill due to repopulation. When
the total dose is zero, the number of cells doubles in a time Tp (potential doubling time
from Section 1.6.4) and the coecient =ln2/Tp. The cells do not begin to proliferate
until Tk (time for kicko from 1.5.4) after treatment begins so the survival becomes:
S=e
nd(+d)+ ln2(TTp Tk )
(1.12)
With the BED becoming (Fowler, 1989):
d
BED = nd 1 +
=
!
ln2 (T Tk )
Tp
(1.13)
1.6.6 Hypofractionation
In the last decade, many groups have reported an / ratio for prostate cancer of
much lower than the value of 10 used for other tumours (Brenner & Hall, 1999; Fowler
et al. , 2001; Brenner et al. , 2002) Values as low as 1.5 have been reported, which
40
1.6. Radiobiological modelling and optimisation
is lower than the value of 3 for late responding tissue (Brenner & Hall, 1999; Fowler
et al. , 2001; King & Fowler, 2001; Kupelian et al. , 2001; Brenner et al. , 2002;
Chappell et al. , 2004). This has implications on fractionation for prostate cancer
radiotherapy. A prostate / ratio lower than that for late responding rectal tissue
means that the prostate BED may always be higher than that for late responding
rectal tissue. It is for this reason hypofractionation will be advantageous for prostate
radiotherapy. Hypofractionation is the delivery of fewer but larger fractions than a
standard fractionation regime, standard being in 2Gy fractions. Hypofractionation
allows the designer of a fractionation regime to achieve one of two things. The rst
is a reduction in late rectal BED for an equivalent prostate BED as compared to a
standard fractionation regime. The second is a greater prostate BED for equivalent late
rectal BED as compared to a standard fractionation regime. The following example
explains this concept. Recall that the BED is dened as:
d
BED = nd +
=
!
(1.14)
For a standard fractionation regime of 78Gy in 39 fractions of 2Gy, this results in a
prostate BED (/ = 1.5) of 182Gy. The late rectal BED (/ = 3) is 130Gy. For an
equivalent prostate BED (182Gy) using a hypofractionated regime of 53.72Gy delivered
in 15 fractions of 3.58Gy, the late rectal BED is 117.86Gy with an SED of 71Gy.
Conversely, for an equivalent rectal BED (130Gy) with a hypofractionated regime of
57.23Gy in 15 fractions of 3.82Gy the prostate BED becomes 202.77Gy with an SED
of 86.9Gy. Many institutions have started implementing hypofractionated regimes for
external beam prostate cancer with some comparable early outcomes (Kupelian et al.
, 2001; Logue et al. , 2001; Kupelian et al. , 2007; Ritter, 2008).
Although there is published data suggesting that the / for the prostate is lower
than that for late responding tissue, this must be taken in the context that there
1.6. Radiobiological modelling and optimisation
41
exists some reports of higher / for the prostate (Nahum et al. , 2003; Wang & Li,
2003). The debate on the subject is outlined in detail in a recent point/counterpoint
discussion (Fowler et al. , 2006).
1.6.7 Tumour Control Probability and Normal Tissue Complication Probability
Tumour Control Probability (TCP) and Normal Tissue Complication Probability
(NTCP) are parameters that describe the probability of specic biological endpoints
(destruction of all tumour clonogens and normal tissue toxicity respectively) as a result
of irradiation. TCP and NTCP are derived from the sigmoid shape of dose response
curves for tumours and normal tissues. Combining TCP and NTCP results in the
probability of uncomplicated tumour control, P+. Figure 1.4 shows example TCP,
NTCP and P+ curves. The sigmoid shape is visible for the TCP and NTCP curves.
There are various mathematical functions that can be used to describe the sigmoid
shape, however poisson and logistic models are most frequently used (Bentzen, 2009).
Tumour and normal tissue response has been described using the Functional SubUnit (FSU) model (Tome, 2009). In this model, tumours and organs are made up of
a number of FSUs. For a tumour, an FSU is a single tumour clonogenic cell. For an
organ, an FSU can be the physiological element of the organ required for function.
For example, nephrons in the kidney or alveoli in the lung or a sub-volume of the total
organ. The FSU model allows description of tumours and organs in terms of their
architecture. Serial architecture is where if only one FSU destroyed, the rest of the
organ is damaged, that is, toxicity is seen when one or more FSUs are destroyed. An
example of a serial organ is the spinal cord. Parallel architecture is where FSUs are
independent of each other therefore death of one FSU does not necessarily damage the
whole organ. Damage to a large number of FSUs is required to aect the function of
42
1.6. Radiobiological modelling and optimisation
Figure 1.4: TCP, NTCP and P+ curves showing the sigmoid shape of the dose-response
curves
the organ. Tumours have perfectly parallel architecture as all FSUs must be destroyed
in order to destroy the tumour.
Dose response curves characterise the probability of a biological end point occurring
for a given dose and can be distinguished by two parameters, the position of the curve
on the dose scale and the slope of the curve. The position of the slope on the dose
scale is represented most commonly as the TCD (for tumour control) or the D50
(for normal tissue) parameters. That is, the dose required, if delivered uniformly to
the organ/tumour, that results in a 50% probability of tumour control or toxicity end
point occurring. The slope of the curve is generally represented by the dose response
gradient, (Bentzen & Tucker, 1997; Brahme, 1984). The dose response gradient
is the change in the probability (in percent) of a given biological end point per one
percent dose increase. The dose response gradient is a function of where it is taken
on the dose response curve; commonly (taken at 37% probability level for poisson
model) and (taken at 50% probability level for logistic model) which is where the
50
37
50
50
43
1.6. Radiobiological modelling and optimisation
model reaches maximum steepness for each model. There values of TCD /D50 and
vary depending on the organ, irradiation conditions and individual person.
A number of NTCP models have been introduced to model the dose response
relationship for normal tissues. Two of the most common methods are the Lyman and
the relative seriality (also referred to as the 'Kallman-S model') models (Lyman, 1985;
Kallman et al. , 1992).
The relative seriality model is based on poisson statistics and assumes that organs
consist of a number parallel subunits, each of which is made up of a number of serial subunits. The relative seriality is represented by the seriality parameter s. The
parameter s is derived from the number of serial subunits in the whole organ. The
NTCP using the relative seriality model is given by:
50
NT CP
(
= 1
voxels
Y
i=1
[1 (P (Di)) ]
sv
)1=s
50
(1.15)
Where v is the relative voxel volume and P(Di) is the poisson dose-response relationship:
P (Di ) = 2
ee (1
(1.16)
D=D50 )
The relative seriality model is thus used with the parameters s, and D50. The
relative seriality model has some drawbacks, in that for a parallel organ, a substantial
NTCP is only calculated if all of the subunits are likely to be damaged (Moiseenko
et al. , 2005).
The Lyman model is one of the most widely used NTCP models:
Z u D;V
p1 1 exp 21 x dx
(1.17)
2
Where D is the absorbed dose and V is the partial organ volume. The upper limit
NT CP (D; V ) =
(
)
2
44
1.6. Radiobiological modelling and optimisation
of the integral:
u(D; V ) =
D D50 (V )
m D50 (V )
(1.18)
Calculation of the NTCP using the Lyman model depends on the model parameters
m, D and n. The parameter m describes the slope of the dose response curve, the
parameter n is the volume parameter (as n increases, the volume eect increases,
that is, tends towards a more parallel architecture) and D is the uniform total dose
necessary for a 50% probability of the specic toxicity end point (that is, D describes
location of the curve on the dose axis). The NTCP using the lyman model also requires
the value of D. The Lyman model assumes D is a uniform dose delivered to the organ,
which is essentially never achieved in radiotherapy practice. The dose to an organ
follows a distribution described by the dose volume histogram (DVH). To calculate
the NTCP using the Lyman model and an organ's DVH, the DVH must be reduced
into a single dose value. The eective volume method is used to achieve this reduction
(Kutcher & Burman, 1989); the DVH is reduced into an equivalent fractional volume
receiving the maximum dose in the DVH:
50
50
50
vEF F
=
X
i
Di 1=n
vi
Dmax
(1.19)
The DVH can also be reduced into the Equivalent Uniform Dose (EUD), which is
the homogenous dose that results in the same biological eect as the heterogeneous
dose (Niemierko, 1997). Calculation of the EUD is covered in the next section. The
upper limit of the NTCP integral thus can be calculated using the EUD:
u(D; V ) =
EUDa D50
m D50
(1.20)
The Lyman model with the EUD reduction has been shown to be mathematically
1.6. Radiobiological modelling and optimisation
45
equivalent to the Lyman model with the Kutcher-Burman reduction (Rancati et al.
, 2004). Throughout this thesis the Lyman model with the EUD reduction will be
referred to as the LKB model.
NTCP calculations provide a probability of a given biological end point. That
is, they do not predict the outcome for a given patient, but provide a complication
probability which can be used to rank plans. There are some disadvantages to the
NTCP models described.
Firstly, NTCP models are based purely on the DVH and don't take into account
the spatial distribution of dose. For organs such as lung, it has been shown that
the eect on lung function depends on where the dose is deposited within the lung
(Bradley et al. , 2007).
Secondly, derivation of NTCP model parameters is performed based on on clinical toxicity rates. Toxicity rates are dependent on the treatment technique, organ
voluming and the genetics of the population analysed. As a result, model parameters
derived from a given clinical data set may not be applicable to another independent
data set (Bradley et al. , 2007).
Lastly, NTCP models are based purely on the dose response function and do not
have any basis on real biological phenomenon. For these reasons, it is suggested that
NTCP models should not be used for clinical decision making, but purely as a research
tool (Bentzen, 2009).
1.6.8 Equivalent Uniform Dose
The Equivalent Uniform Dose (EUD) parameter was developed to provide a means of
equating heterogeneous dose distributions using a single parameter (Niemierko, 1997).
The EUD model includes both physical dose and a biological parameter based on the
seriality of the organ in question. The EUD equation:
46
1.7. Measurement modalities
EUD =
X
i
vi Dia
!1=a
(1.21)
The calculation of EUD is based on the volume of the volume element of interest,
vi , the physical dose to that volume element, Di and the parameter a, which is equal
to the value of 1/n from the LKB NTCP model. The behaviour of the EUD function
is dependent on the parameter a, in that a is used to describe whether the volume
is a tumour or normal tissue, and the seriality of the volume. A negative a value is
used for tumours and a positive a value is used for normal tissue. For normal tissue,
as a approaches innity, EUD approaches the maximum dose. Large values of a are
used for serial organs. For an a value of one, the EUD is equal to the mean dose to
the structure. As a approaches negative innity EUD approaches the minimum dose.
Tumours are the ultimate parallel organ, that is, they require all cells in the structure
to be killed for the tumour to be killed. For this reason, a highly negative value of a is
selected since the lowest dose to the tumour denes the control probability (Wu et al.
, 2002).
EUD was originally introduced as a means of comparing dose distributions, and is
still used for this. However, recent use of EUD has been as an optimisation function for
IMRT plans. When optimising IMRT plans based on the gradient reduction scheme, it
is useful if the optimisation function can be easily dierentiated. The EUD is one such
function. A detailed description of the EUD function and its use in IMRT optimisation
is given in Chapter 3.
1.7 Measurement modalities
Various dosimeters are used clinically for radiation dose measurement. The 'gold
standard' detector, the ionisation chamber, is most commonly used for beam commis-
1.7. Measurement modalities
47
sioning and quality assurance. However, the ionisation chamber only provides a point
dose measurement and its volume can be too large for some applications. In particular,
high dose gradients in the build up region and penumbra may require better spatial
resolution than some ion chambers can provide. Therefore other detectors are used,
such as lm and solid state detectors, for non-routine applications.
1.7.1 Ionisation chambers
Ionisation chambers (ion chambers) are the most common and reliable dosimeters
used in radiation oncology applications. Ion chambers consist of detecting ionisation
processes in a gas and relating them to a given absorbed dose. When charged particles
traverse a cavity of gas, molecules are ionised resulting in a positively charged ion and a
free election (an ion pair). In an ion chamber, a gas cavity is placed between electrodes.
Air is commonly used as the gas. A bias voltage is applied across the electrodes.
When ion pairs are created by charged particles traversing the gas chamber, positively
charged ions and the free electrons move away from their point of origin towards the
electrodes. This drift of ions and electrons constitutes an electric current and charge
is collected. The collected charge is measured by an electrometer. Various correction
factors and a calibration factor converting from units of charge (Coulombs) to absorbed
dose (J/Kg or Gy) are applied to the charge reading to obtain the absorbed dose.
The ion pairs created during ionisation in the chamber are subject to recombination. Therefore, a suciently high bias voltage (> 100V) must be applied to reduce
the number ion pairs recombining and increase the charge collection, whilst too high
bias voltage (> 400V) results in ions in the chamber gaining sucient energy to cause
secondary ionisation. Further increasing of the bias voltage results in avalanches of
ions per primary photon (the Geiger-Muller eect), which results in a much higher
sensitivity. For radiotherapy applications, ion chambers are operated in the plateau
1.7. Measurement modalities
48
region between 100V and 400V.
1.7.2 Radiographic lm
Radiographic lm for use in radiation oncology consists of a radiosensitive silver halide
emulsion on a polyester base. The silver halide emulsion commonly consists of silver
bromide crystals embedded in gelatine. When photons are interact with the silver
bromide, the Silver Bromide molecules are ionised resulting in Bromide and an electron.
The electron combines with the positive Silver Ion resulting in elemental silver. When
the lm is developed all the Silver ions are reduced in any crystal that contains a
reduced Silver ion. Silver halide crystals that don't contain reduced Silver ions are
washed away, leaving only reduced Silver ions, coloured black. In other words, parts
of the lm that are irradiated turn black.
The density of the lm, that is, the darkness of the lm is proportional to the
absorbed dose in the lm is measured using either spot densitometers or specially
designed transmission lm scanners such as the Vidar (VIDAR Systems Corporation,
VA, USA) scanner. Radiographic lm provides an accurate high resolution 2D dose
map and as a result is used extensively in quality assurance in radiotherapy. A disadvantage of radiographic lm is its high eective atomic number which leads to an over
response to low energy x-rays. Radiographic lm is sensitive to visible light therefore it
is packaged in a light-tight envelope, the integrity of which must be maintained during
use until processing. Radiographic lm also requires a well-maintained lm processor.
There are some issues with parallel exposure (Suchowerska et al. , 2001).
Radiographic lm has been used extensively in the past and is still used clinically however energy response issues and the reduction in the use of lm developers
suggests that radiochromic lm will become the more common lm for radiotherapy
applications.
1.7. Measurement modalities
49
1.7.3 Radiochromic lm
A relatively new dosimeter to the radiotherapy dosimetry setting is radiochromic lm.
Radiochromic lm contains a chemical structure that changes colour when exposed to
ionising radiation. This means radiochromic lm is a self-developing lm that doesn't
require any post-processing to obtain an image. Radiochromic lm is also relatively
insensitive to visible white light; it doesn't need to be kept in a light tight package
while in use. Radiochromic lm turns blue when exposed to radiation due to a solidstate polymerisation in which coloured, polyconjugated polymer chains are created
(Niroomand-Rad et al. , 1998). The shade of blue is proportional to the absorbed
radiation dose.
The optical density of radiochromic lm can be read out using atbed scanners,
spot densitometers or transmission scanners however atbed scanners are the recommended read out method. Advantages of radiochromic lm include:
No processing required
Can be exposed to visible white light during use
Can easily be cut to any desired shape
High spatial resolution
Instantaneous image is obtained
Can be immersed in water
The radiochromic lm used in this thesis is Gafchromic EBT lm (International
Specialty Products, Wayne, NJ, USA). EBT lm is specically designed for radiotherapy applications with a sensitive dose range of 0.1-800cGy. The structure of EBT
lm is shown in Figure 1.5. EBT lm does not contain any high Z materials, like
radiographic lm, therefore the eective atomic number of EBT lm is close to that
1.7. Measurement modalities
50
Figure 1.5: The structure of Gafchromic EBT lm (ISP, 2007)
of water, at Z=6.98 (ISP, 2007). A more recent version of radiochromic lm - EBT-2
- has just been released (ISP, 2009). EBT-2 lm diers from EBT lm in that it is
yellow in colour and has a dierent geometry. EBT-2 lm was not used in this thesis.
Radiochromic lm has been the subject of various reports describing a number of
processes that must be carried out when analysing the lm to obtain highly accurate
measurements. These include preservation of scanning orientation, not using the edge
of cut lms for measurement, keeping post-irradiation readout time to > 24 hours and
storage and handling conditions (Niroomand-Rad et al. , 1998; Cheung et al. , 2005;
Yu et al. , 2006).
1.7.4 Metal Oxide Semiconductor Field Eect Transistor detectors
Metal Oxide Semiconductor Field Eect Transistor (MOSFET) detectors are solid
state radiation detectors that are used for point dose measurements requiring high
spatial resolution. MOSFETs as radiation detectors were rst proposed in 1974 by
Holmes-Siedle (1974) and have been used extensively for radiotherapy applications
(Rosenfeld, 2002). MOSFET radiation detectors are simply conventional MOSFET
51
1.8. Disequilibrium region dosimetry
Figure 1.6: Schematic diagram of a MOSFET radiation detector
chips that have a thicker semiconductor substrate. The structure of a typical MOSFET
detector is given in Figure 1.6.
Ionising radiation incident on a MOSFET detector interacts in the SiO layer.
Electron-hole pairs are created. Electrons move towards the gate and positive holes
move towards the Si-SiO interface and get trapped. This causes a positive charge
build up. The eect of the charge build up is to alter the threshold voltage, Vt. The
threshold voltage, in MOSFET radiation dosimetery, is dened as the voltage at which
a given current ows. The MOSFET is then 'read out' by putting a constant current
through the drain-source and measuring the Vt for that given current. The change in
Vt is proportional to the absorbed dose.
MOSFET detectors can be used in passive or active mode. In passive mode, no bias
is applied to the gate electrode during irradiation. In active mode, a bias is applied
to the gate electrode during irradiation. This reduces electron-hole recombination by
acting to separate electron-hole pairs, increasing the sensitivity of the detector.
2
2
1.8 Disequilibrium region dosimetry
Disequilibrium region (DR) dosimetry is dened here as dosimetry in any location
where electronic equilibrium does not exist. In this work, DR dosimetry refers to
52
1.8. Disequilibrium region dosimetry
dosimetry on patient or phantom surface and dosimetry in the regions surrounding air
cavities in a patient or phantom.
The measurement of DR dose is a complicated task. Many factors aect the measurement such as depth of measurement and perturbation of the radiation due to
the detector used. Monte Carlo techniques have been employed as well as solid state
dosimeters (such as MOSFET detectors), radiographic and radiochromic lm and thermoluminescent dosimeters (TLDs).
When considering DR dosimetry, the depth of measurement is extremely important, particularly for skin dosimetry. The ICRP recommends the depth of skin dosimetry to be at the average depth of the basal layer in the skin, (the basal layer being the
radiosensitive layer of the skin) (ICRP, 1991). This depth can vary from 20m (trunk
and face) to up to 560m (ngertips). For most parts of the skin, the depth of the
basal layer is found between 50m and 100m therefore the depth of measurement
recommended by the ICRP is 70m (ICRP 1991).
Figure 1.7 shows a depth dose curve for a 6MV, 10x10cm photon beam from a
linac incident on a water block phantom. Only the rst 1.5cm depth is presented,
showing the high dose gradients found at the surface of a phantom or patient. Figure
1.7 illustrates that any small variation in detector depth will yield a dierent dose
measurement. Any discussion of 'skin dose', 'surface dose' or 'supercial dose' must
acknowledge the depth of measurement so that comparisons can be made with other
data sets.
Each detector has an intrinsic build up due to its construction. Therefore, when
discussing DR dosimetry, the term water equivalent depth (WED) is used. The WED
of a detector location is the equivalent depth if the build up material on the detector
was made from water.
Monte Carlo techniques are attractive for DR dosimetry due to the highly cus2
1.8. Disequilibrium region dosimetry
53
Figure 1.7: A 6MV depth dose curve for the rst 1.5cm depth in water showing the
steep dose gradient at the surface. The curve was generated using the BEAMnrc/DOSXYZnrc Monte Carlo package using a voxel resolution of 100m in the depth
direction
tomisable resolution that can be obtained and the fact that no perturbation of the
beam takes place. However, as resolution increases, the number of histories required
for statistical accuracy increases. Monte Carlo allows for DR dosimetry that removes
any systematic errors (such as detector perturbation) at the expence of stochastic
(random, statistical) errors. However, the accuracy of radiotherapy DR dosimetry
with Monte Carlo is highly dependent on the accuracy of the phase space data used
for simulation (the le describing the incident radiation). For example, for surface
dosimetry the electron contamination must be modelled accurately.
Parallel plate ion chambers provide very accurate DR measurements and are commonly used as a 'gold standard' in surface dosimetry. Parallel plate chambers such as
'Attix chambers' consist of a cylindrical chamber placed on the surface of the phantom
or at an interface region with the central axis of the cylinder in the same direction as
1.8. Disequilibrium region dosimetry
54
the beam. Attix chambers provide accurate measurements at the surface of a phantom
or cavity i.e. 0m (Rawlinson et al. , 1992). Attix chambers can be quite cumbersome
to use, as they must be embedded in the phantom surface. Their relatively large size
(they can be 2-3cm wide) limits the use of Attix chambers outside of standard block
phantom geometry. Parallel plate ion chambers require some correction for an overresponse caused by ionisation in the chamber from electrons created in the walls of
the chamber. This has been described by Rawlinson et al. (1992).
DR dosimetry with MOSFETs and other semiconductor detectors such as diodes
provides very high depth resolution. The sensitive volume of MOSFET detectors is
the thin SiO layer which is typically 1m thick. This is much less than what is
achievable with other detectors. The construction of semiconductor detectors is such
that perturbation of the beam occurs. Detector packaging can non-uniformly attenuate
the beam prior to interaction with the sensitive volume leading to angular dependence.
Commercial MOSFET detectors have their sensitive detection volume at a WED of
between 0.7mm and 1.8mm (Butson et al. , 1996; Scalchi et al. , 2005).
Radiographic and radiochromic lms have been used to great eect in DR dosimetry (Devic et al. , 2006). Placed at an interface region, a lm allows for a 2D surface
dose map to be obtained. Radiochromic lm is the preferable DR dosimeter due to its
atomic number similar to that of water, and the fact that it can be cut and shaped easily to t almost any geometry (radiographic lm must be contained in a light-tight envelope during measurement). A common radiotherapy radiochromic lm, Gafchromic
EBT (International Specialty Products, Wayne, NJ, USA) has a 40m thick active
layer centred at a WED of 153m (Devic et al. , 2006).
2
Chapter 2
Rectal dose reduction with IMRT
for prostate cancer radiotherapy
2.1 Introduction
Clinicians are still seeking the optimal planning method when treating prostate cancer
with external beam radiotherapy. It is well established that rectal dose and late rectal
toxicity are correlated (Boersma et al. , 1998; Skwarchuk et al. , 2000; Jackson et al.
, 2001). Numerous comparisons of dierent methods have been undertaken utilising
DVH parameters to determine the superiority of any particular technique (Bedford
et al. , 1999; Fiorino et al. , 2000; Khoo et al. , 2000). The corollary of these eorts
has been a well dened reduction in acute and late rectal toxicity despite signicant
dose escalation from doses of 60-68Gy with a 'conventional' four eld box to 74-78Gy
non-coplanar, image-based, conformal arrangements (Livsey et al. , 2003; Cozzarini
et al. , 2007).
Part
of this chapter has been submitted for publication in the Journal of Medical Imaging and
Radiation Oncology:
Hardcastle, N., Davies, A., Foo, K., Miller, A. and Metcalfe, P. E. (2009). Rectal Dose Reduction
with IMRT for Prostate Cancer Radiotherapy. Journal of Medical Imaging and Radiation Oncology
(in submission)
55
2.2. Method and materials
56
There is some evidence to suggest that for prescribed doses of up to 74Gy, a three
eld approach is suitable, but further dose escalation delivers doses of >50Gy to the
femoral necks (Khoo et al. , 2000). Although a higher incidence of hip fractures has
not been reported with these doses, few oncologists are prepared to exceed previously
reported tolerance doses (Emami et al. , 1991). Further, even modest dose escalation
above 70Gy can increase rectal toxicity signicantly (Dearnaley et al. , 2007). This
has led to the introduction of many multi-eld 3D Conformal Radiation Therapy
(3DCRT) and Intensity Modulated Radiation Therapy (IMRT) approaches to solve
the problem of delivering an extremely high prostate dose while limiting the dose
delivered to the immediately adjacent rectum. The use of IMRT in localised prostate
cancer radiotherapy has been shown elsewhere to lead to reduced rectal toxicity for
equivalent or higher target dose (Zelefsky et al. , 2000, 2001; Kupelian et al. , 2002a,b;
Namiki et al. , 2006; Sanguineti et al. , 2006; Veldeman et al. , 2008). Despite this
evidence, IMRT for prostate radiotherapy is not standard in the Australian clinical
setting.
A systematic approach has been taken to compare IMRT plans with the current clinically standard (at Illawarra Cancer Care Centre (ICCC)) 3DCRT plans for
prostate radiotherapy. Dosimetric and biological quantities as well as delivery eciency are compared in an eort to quantify any dosimetric and radiobiological dierences between 3DCRT and IMRT.
2.2 Method and materials
All plans were generated on a Pinnacle Radiotherapy Planning System V7.6c with
the P IMRT optimisation toolbox (Philips Radiation Oncology Systems, Fitchburg,
WI, USA). Sixteen sequential clinical patients were chosen for analysis. None of the
patients had seminal vesicle (SV) involvement. Patients for which the prostate only
3
2.2. Method and materials
57
was the target were chosen as they represent the majority of the prostate radiotherapy
cases at the ICCC. It is acknowledged that IMRT may have a greater advantage over
3DCRT when SVs are included in the PTV and this is discussed in Section 10.9. All
of these patients received ve-eld conformal radiotherapy. The ve-eld conformal
plan will be referred to as 3DCRT in the remainder of the report.
The clinical target volume (CTV) was dened as the prostate, volumed by a single
radiation oncologist for all cases. While all of the simulation CTs had volumes and
contours marked according to the Trans-Tasman Radiation Oncology Group (TROG)
RADAR trial protocol (TROG 03.04), none of the patients were entered on this TROG
protocol. The CTV conformed to the visible extent of the prostate on the non-contrast
CT scan. The inferior extent of the prostate was determined on sagittal reconstruction
(CT slice width of 2mm), and seminal vesicles were not contoured once separated from
the prostate gland. No MRI fusion was employed.
A single 7mm CTV to planning target volume (PTV) 3D expansion was applied,
and a prescription dose of 78Gy delivered to the isocentre in 39 fractions was set
(Pollack et al. , 2000; Kuban et al. , 2008). Whilst a two phase technique with reduced
posterior margin is often used (Skala et al. , 2004), our institution uses a single target
volume. This approach is consistent with symmetrical setup uncertainty and does not
increase rectal dose Davies et al. (2008a). It also makes for a more robust comparison
with IMRT techniques to simplify analysis. The rectal volume was dened using full
volume and the length from anal canal up to the anterior curve of rectum into sigmoid
colon with a maximum length of 11cm. The rectum can be dened as the rectal wall,
that is, a 2-3mm thick ring dened by the external contour of the rectum, as well as
the solid rectal volume. The eect of the two rectal denition methods, solid rectum
or rectal wall only, on the resultant DVHs was investigated by comparing the average
rectal DVHs for the 3DCRT and IMRT plans. The rectal wall contour was derived
2.2. Method and materials
58
from the solid rectum contour by creating a uniform 3mm thick ring structure based
on the external contour of the solid rectum .
2.2.1 3DCRT plan
The 3DCRT plans were planned by a single experienced radiation therapist and were
used clinically to treat the patients. The beam angles were 90, 45, 0, 315 and
270 (IEC, 1996). This technique has been found in our institution to be superior
to a six eld approach Davies et al. (2008a). Wedges were used for all elds except
the 0 eld. Beam weightings and wedges were adjusted to individually optimise plan
coverage but typically the lateral elds (90 and 270) used 45 wedges, while the
anterior oblique elds (45 and 315) used 25 wedges. Beam weighting were evenly
distributed initially before adjusting to achieve an optimal forward-plan distribution.
Beam angles were not adjusted. Multileaf Collimator (MLC) shielding was used with
a leaf width at isocentre of 5mm. The MLC aperture was set to a 6mm margin around
the PTV.
2.2.2 IMRT plan
A seven-eld IMRT plan was generated for each patient, referred to as 'IMRT'. All
plans had elds at gantry angles 120, 80, 40, 0, 320, 280 and 240 (IEC, 1996). All
IMRT plans were optimised using Direct Parameter Machine Optimisation (DMPO)
which is a form of direct aperture optimisation (Shepard et al. , 2002). The optimal
uence map is derived using only apertures that are deliverable by the collimator. A
maximum of 70 segments were used for each plan and 50-75 iterations were run for
each plan to provide an optimal solution using the weighted gradient minimisation
method.
The coverage on the PTV was matched to the 3DCRT plan. In the clinical setting,
2.2. Method and materials
59
Table 2.1: IMRT optimisation parameters. ROI = Region Of Interest, DVH = Dose
Volume Histogram, ALAP = As Low As Possible
ROI
Objective type Dose (Gy) % Volume Weight
min dose
74.1
constraint
PTV
max dose
81.0
constraint
min DVH
76.0
95
100
max dose
76.0
constraint
max DVH
25
ALAP
30
Rectum
max DVH
50
ALAP
30
max DVH
60
ALAP
30
max DVH
70
ALAP
30
Bladder
max DVH
50
ALAP
30
Femoral Heads max dose
50
30
Ring
max DVH
74.1
5
25
coverage of the PTV is the primary optimisation objective, following which minimisation of at-risk organ doses can and should occur. Although slight improvement on
the PTV coverage could have been obtained with the IMRT plans, the IMRT plans
should not be subject to more strict PTV coverage objectives than the 3DCRT plans
if improvements in rectal DVHs are being investigated. For the IMRT plan, the rectal DVH was optimised using a maximum dose constraint and a series of maximum
DVH objectives. The maximum DVH objectives were set at 25Gy, 50Gy, 60Gy and
70Gy, with incremental reduction as PTV constraints continued to be met so that the
volumes achieved were as low as possible (ALAP) without compromising the PTV
coverage.
The maximum DVH objective volumes varied from patient to patient and required
constant monitoring and updating such that the rectum was receiving the lowest possible doses without impacting on the PTV coverage. The maximum volumes used in
the maximum DVH objectives depend on the specic patient's anatomy therefore it
was necessary that these were tailored to each patient. It is acknowledged that there
are many variables in IMRT optimisation such as segment numbers, minimum segment
2.2. Method and materials
60
Figure 2.1: Dose distributions for patients #7 and #11. The left image shows the
3DCRT plan and the right image shows the IMRT plan. The dose scale ranges from
0-80Gy.
sizes and monitor units, objective types and number of iterations that can impact on
plan quality. In addition to the target and organs at risk objectives, a ring structure
was dened as a 1.5cm thick ring around the PTV. This is a structure used purely
for optimisation purposes to increase the dose gradient away from the target. The
optimisation objectives are given in Table 2.1.
2.2.3 Evaluation of results
The CT data sets, contours and calculated dose distributions were exported into the
Computational Environment for Radiotherapy Research platform (Deasy et al. , 2003)
(CERR, Washington University in St. Louis, St. Louis, MO, USA) and a script was
2.3. Results
61
written in MATLAB to export and calculate average target and organ at risk DVHs
over all 16 patients for each plan. In addition to this the normal tissue complication
probability (NTCP) for the rectal Region of Interest (ROI) was calculated using the
Lyman-Kutcher-Burman (LKB) model (Lyman, 1985; Kutcher & Burman, 1989). The
LKB calculation utility in CERR was used. The choice of LKB model parameters has
a large eect on the resultant NTCPs. Three sets of model parameters representing
rectal bleeding of grade 2 were used so to take into account the eect of the model
parameters on the plan ranking. The rst parameter set, NTCP1 (n=1.03, m=0.16
and D50=55.9Gy) (Tucker et al. , 2004a), the second parameter set NTCP2 (n=0.24,
m=0.14 and D50=75.7Gy) (Rancati et al. , 2004), and the third parameter set NTCP3
(n=0.084, m=0.108 and D50=78.4Gy) (Sohn et al. , 2007), were taken from the literature. The number of monitor units required for each plan was recorded for comparison
of the delivery eciency of each plan. The Wilcoxon matched-pair signed-rank test
was used to compare the above parameters. The threshold for statistical signicance
was p < 0.05.
2.3 Results
2.3.1 Dose-volume comparison
Table 2.2 shows close agreement between the maximum, minumum, mean and standard
deviations of the average PTV doses between the 3DCRT and IMRT plans. The IMRT
plans had a slightly increased standard deviation of the PTV dose which suggests a
more heterogeneous dose distribution with the IMRT plans, similar to that found by
others (Vaarkamp et al. , 2009; Luxton et al. , 2004). The dose distributions for the
3DCRT and IMRT plans for two selected patients are given in Figure 2.1.The resultant
average DVHs are shown in Figure 2.2. The individual patient rectal and PTV DVHs
2.3. Results
62
Figure 2.2: Average cumulative DVHs for (a) PTV and Rectum and (b) Femoral
Heads and Bladder. The individual patient DVHs can be found in Figure 2.3
2.3. Results
63
Figure 2.3: Individual patient PTV and Rectal cumulative DVHs for all patients in
the study
2.3. Results
64
are given in Figure 2.3. Figure 2.2a shows the cumulative DVH data for the rectum
and PTV for the 3DCRT and IMRT plans. The 3DCRT and IMRT plans both have
similar coverage of the PTV. The IMRT plans have reduced rectal volumes receiving
doses > 5Gy. Figure 2.2b shows the cumulative DVH data for the bladder and femoral
heads for the 3DCRT and IMRT plans. The IMRT plans show a small dose-volume
reduction over the 3DCRT plan for the bladder. IMRT plans show a large dose-volume
reduction compared with the 3DCRT plan in the femoral heads. Table 2.3 gives the
average rectal V25Gy, V50Gy, V50Gy, V70Gy and V75Gy values for the 3DCRT and
IMRT plans. The IMRT plans resulted in statistically signicant lower values for all
ve parameters.
Figure 2.4a and Figure 2.4b show the average DVHs over the 16 patients of the solid
rectum and rectal wall contours for the 3DCRT and IMRT plans respectively. The
average V25Gy-V75Gy values are given in Table 2.4. For both planning techniques,
small, but statistically signicant dierences between the solid rectum and rectal wall
contour denition were observed for both 3DCRT and IMRT techniques. For the
3DCRT plans the solid rectum contour had a larger average V25Gy but smaller V60,
V70 and V75Gy values. For the IMRT plans a statistically signicant dierence was
seen only for V50, V60, V70 and V75Gy where the solid rectum was lower than the
rectal wall contour.
The selected patients shown in Figure 2.1 are presented as they represent maximum
and minimum rectal DVH gains. From Figure 2.3 it is seen that for Patient 7, minimal
rectal DVH reduction is observed for IMRT over 3DCRT. In this particular case, the
rectum extends posteriorly away from the target by a large distance (Figure 2.1). This
results in a large proportion of the rectal volume being away from the target hence
out of the way of the beams. For this particular anatomy, gains made by IMRT will
be minimal as the 3DCRT can already achieve signicant rectal sparing. In patient
2.3. Results
65
Table 2.2: PTV coverage metrics (averaged over all 16 patients)
dose maximum dose mean dose standard
plan minimum
dose (Gy)
dose (Gy) dose (Gy) deviation (Gy)
3DCRT
72.3
80.5
77.8
1.3
IMRT
72.6
81.8
77.8
1.4
p-value
not sig.
< 0.001
not sig. 0.05 < p < 0.10
Table 2.3: Average rectal percentage volumes receiving 25, 50, 60, 70 and 75Gy
Parameter 3DCRT IMRT
P value
V25Gy 65.29 52.81
< 0.001
V50Gy 28.81 26.76 0.01 < p < 0.02
V60Gy 23.99 22.23 0.02 < p < 0.05
V70Gy 17.38 15.10 0.005 < p < 0.01
V75Gy 10.68 7.54
< 0.001
7, there is a large amount of gas in the rectum, which allows sparing of the rectum.
This is similar to that achieved articially with rectal balloon, in that the posterior
rectal wall is forced away from the target area. Conversely, for Patient 11, reductions
in the rectal DVH are made over the whole dose range. In this case, the rectal volume
is small (Figure 2.1) and a larger proportion of the rectal volume is close to or within
the target volume, so any modulation of the beams has a greater eect at reducing
the dose to the organ.
2.3.2 Biological parameter comparison
Figure 2.5 shows the normal tissue complication probability (NTCP) for the rectal
DVH for all patients. Figure 2.5a, b and c shows the NTCP data for all patients using
the NTCP1, NTCP2 and NTCP3 parameter sets respectively. Figure 2.5a and Figure
2.5b show that for NTCP1 and NTCP2 respectively, IMRT resulted in a lower rectal
NTCP for all patients. Figure 2.5c shows that for NTCP3, IMRT resulted in a lower
rectal NTCP for all but one patient. The average NTCP and statistical signicance is
2.3. Results
66
Figure 2.4: Average Solid rectal DVH vs rectal wall DVH for a) 3DCRT and b) IMRT
plans
2.3. Results
67
Table 2.4: V25Gy - V75Gy parameter values for Solid Rectum (SR) and Rectal Wall
(RW) contours for 3DCRT and IMRT plans
Parameter Contour 3DCRT
p-value
IMRT
p-value
V25Gy
SR
63.0 0.01 < p < 0.02 50.1
> 0.2
RW
59.1
49.3
V50Gy
SR
26.7 0.05 < p < 0.10 24.2 0.02 < p < 0.05
RW
28.2
26.2
V60Gy
SR
20.7 0.01 < p < 0.02 18.2 0.01 < p < 0.02
RW
23.3
20.9
V70Gy
SR
14.4 0.001 < p < 0.005 11.9 0.001 < p < 0.005
RW
17.9
15.3
V75Gy
SR
8.4 0.001 < p < 0.005 5.8
< 0.001
RW
11.9
9.1
Table 2.5: Average NTCP values for 3DCRT and IMRT plans with statistical signicance
Parameter set 3DCRT IMRT p-value
NTCP1
4.61 1.59 < 0.001
NTCP2
2.51 1.69 < 0.001
NTCP3
6.69 5.10 < 0.001
given in Table 2.5.
For the selected patients in Figure 2.1, the rectal NTCPs reect the DVHs of each
plan. For Patient 7, as the relative rectal volumes receiving doses over the whole
dose range are relatively low, compared with other patients, the NTCPs are amongst
the lowest out of all patients in the study. Only for NTCP3, which is related to
the high dose region hence that section of the rectum contained by the PTV, does
the modulation in the IMRT achieve some reduction. For Patient 11, as the DVH
reduction with IMRT is seen over the whole dose range, reduction in NTCP is seen
for all three sets of model parameters.
Figure 2.6 shows the rectal NTCP for each patient plotted against the percentage
of the rectal volume contained by the PTV for the three NTCP parameter sets. Figure
2.6a, Figure 2.6b and Figure 2.6c show the rectal NTCP for parameter sets NTCP1,
2.3. Results
68
NTCP2 and NTCP3 respectively, with the Spearman's rank correlation and p values.
For parameter set one (Figure 2.6a), no statistically signicant correlation is seen between the proportion of the rectal volume contained by the PTV and the rectal NTCP.
Parameter set one has an n value that represents a more parallel organ architecture
hence the high dose region has little impact on the NTCP. For parameter sets NTCP2
and NTCP3, a statistically signicant correlation is seen between the proportion of
the rectal volume contained by the PTV and the NTCP. Parameter sets NTCP2 and
NTCP3 have values of n that represent more serial organ architecture. This means
that any high dose region is penalised. As the proportion of rectal volume contained
by the PTV is receiving the target dose, any increase of this volume increases the
NTCP. As the value of n increases (more serial) the correlation value increases (that
is, the correlation becomes stronger).
For the two parameter sets that resulted in a statistically signicant correlation
between percentage rectal volume contained by the PTV and NTCP (NTCP2 and
NTCP3), a higher value of the correlation coecient was observed for the IMRT plans
over the 3DCRT plans. This suggests that the proportion of the rectum contained by
the PTV is a more dominant predictor of NTCP for the IMRT plans than for the
3DCRT plans. This is probably due to the reductions in rectal volumes irradiated to
high doses observed for the IMRT plans. The overlap region is where the majority of
the high doses are delivered with the IMRT plans. For the 3DCRT plans, the inability
to sculpt the dose means that the overlap region plus extra rectal volume gets irradiated
to high doses. This has the eect of increasing the complication probability for rectal
complications related to high doses.
2.3. Results
69
Figure 2.5: Rectal NTCPs using (a) model parameters n=1.03, m=0.16 and
D50=55.9Gy (b) model parameters n=0.24, m=0.14 and D50=75.7Gy and c) model
parameters n=0.084, m=0.108 and D50=78.4Gy
2.3. Results
70
Figure 2.6: Rectal NTCP vs percentage of rectal volume contained by the PTV
for (a) model parameters n=1.03, m=0.16 and D50=55.9Gy (b) model parameters
n=0.24, m=0.14 and D50=75.7Gy and c) model parameters n=0.084, m=0.108 and
D50=78.4Gy. Spearman's rank correlation coecient and p values are presented on
each chart
2.4. Discussion
71
Table 2.6: Average MU per plan averaged over 16 patients
Plan Average MUs Range P value
3DCRT
376
325-405
IMRT
540
464-658 < 0.001
2.3.3 Delivery eciency comparison
The number of monitor units (MU) required to deliver each plan was recorded. The
average MU per delivery for each plan over the 16 patients is summarised in Table 2.6.
2.4 Discussion
The presented DVH data shows that for the same PTV dose coverage, physically
optimised seven eld IMRT resulted in a reduction of rectal doses compared with a ve
eld 3DCRT plan. This was seen over the whole dose range. Statistically signicant
reductions in V25Gy, V50Gy, V60Gy, V70Gy and V75Gy values were obtained with
IMRT plans. It should be noted, however, that the mean reduction in rectal dose from
using IMRT, although statistically signicant, is small in absolute terms, particularly
for doses 50Gy and higher. An examination of individual patient data in Figure 2.3
shows that, although the average rectal dose reduction is small, some patients would
derive much greater advantage from IMRT than others.
When the rectal wall and solid rectum DVHs were compared it was observed that
for the 3DCRT plans the average rectal wall DVH was larger for low-mid range doses
but lower at the high dose range. For the IMRT plans the average rectal wall DVH
was larger for the mid-high range doses. The dierences between the rectal wall and
solid rectum DVHs were small but statistically signicant. These results are similar
to that found in previous studies (MacKay et al. , 1997; Fenwick et al. , 2001; Tucker
et al. , 2004a).
2.4. Discussion
72
The bladder doses were comparable for all plans; however little emphasis was placed
on reducing the dose to the bladder apart from reducing V50Gy. The greater interfraction variability of bladder positioning, due to variable lling precludes more aggressive
optimisation of bladder dose without daily soft tissue imaging and adaptive radiotherapy planning (McBain et al. , 2009; Fiorino et al. , 2005).
All IMRT plans resulted in a large reduction in femoral head dose. In the same
way that the ve eld approach reduces dose to the femoral heads over a three eld
approach by distributing dose over more elds, so a seven eld approach will also
reduce dose through the lateral elds. One solution for ve eld 3DCRT to reduce
femoral head dose is to rotate the lateral beams posteriorly to the femoral heads, but
this nullies the rectal dose advantage gained by the ve-eld method.
Overall, the dosimetric advantages for IMRT over 3DCRT were real, but small.
This is perhaps surprising, given the evidence for IMRT superiority cited previously
(Zelefsky et al. , 2000, 2001; Kupelian et al. , 2002a,b; Namiki et al. , 2006; Sanguineti
et al. , 2006; Veldeman et al. , 2008). The reason for the smaller than expected
dosimetric gains may lie in the use of a high quality comparator in the 3DCRT plans.
The comparison 3DCRT method was the result of extensive work to nd the best classsolution technique for prostate radiotherapy using forward-planning (Davies et al. ,
2008b). Pushing the technical limits of 3DCRT will naturally reduce the apparent
advantages of IMRT. However, there is a limit to how much 3DCRT can achieve. The
rst plan produced in the forward planned 3DCRT process includes little control over
the resulting dose distribution. Gains are made by iterations based on intuition and
experience. This approach is best suited to meeting a small number of dose-volume
objectives, particularly at the high end of the dose range.
Australian guidelines for dose-volume objectives for the rectum in prostate radiotherapy have concentrated on high dose volumes above 60Gy with the aim being
2.4. Discussion
73
reduction of rectal bleeding (Skala et al. , 2004). However, it is becoming clear that the
moderate dose region also contributes signicantly to rectal toxicity endpoints such as
rectal incontinence (Gulliford et al. , 2009; Peeters et al. , 2006; Fiorino et al. , 2008).
Whilst forward planned 3DCRT is a proven method to meet dose-volume constraints
at high doses, this task becomes exponentially more complex as more constraints are
added at lower dose-volume points. Inverse planned IMRT is better suited to the task,
and also results in gains in the high dose region.
The NTCP comparison shows that IMRT use results in a decrease in modelestimated likelihood of toxicity in the majority of patients. Although the radiobiological model used has its limitations, it has been shown to t toxicity data in a
number of large real-world datasets, albeit with local modications (Peeters et al. ,
2006; Rancati et al. , 2004). The value of this analysis is not that it accurately predicts the actual rate of toxicity, but that it provides a biologically-based comparison of
which plan is less likely to be associated with toxicity, based on currently available evidence. There is wide variation in published values of the LKB model parameters D50,
m and n (Peeters et al. , 2006; Rancati et al. , 2004; Sohn et al. , 2007; Tucker et al.
, 2004a). This variation is due to natural variation in radiosensitivity, heterogeneous
toxicity endpoints and the actual spatial distribution of dose within volumes that not
taken into account by DVH-based models. For each of the three sets of published
model parameters, each dierent, but all based on clinical data, the IMRT plans are
generally superior to 3DCRT. That is, the superior NTCP results from IMRT plans
are robust even in the face of model parameter uncertainty.
All IMRT plans were less MU ecient than the 3DCRT plans, which is expected
due to the delivery technique employing multiple segments per beam, as opposed to a
single segment used per eld for 3DCRT. This relative ineciency raises two possible
concerns. The rst is whether the use of more monitor units and the attendant leakage
2.4. Discussion
74
dose increases the risk of stochastic late eects such as second malignancy, as raised
by Hall (Hall & Wuu, 2003). This is controversial, and may be oset by greater
conformality of higher dose (Tubiana, 2009). The second concern is the time-eciency
of delivery, where a greater number of MU directly increases delivery time for each
beam. In addition to this, for step and shoot IMRT, the delay between segment
delivery due to leaf motion slightly increases delivery time. It can also be argued
that a signicant time resource increase for IMRT results from the required quality
assurance procedures.
It may be possible to reduce the required MU for the IMRT plans by reducing
the maximum number of segments available to the optimiser. Reducing the number of
segments may reduce the dosimetric advantages obtained with IMRT; however, this has
not been formally tested in this report. Furthermore, new IMRT delivery techniques
such as Volumetric Modulated Arc Therapy (VMAT) have been shown to result in
similar dose coverage to conventional IMRT, but can be delivered using signicantly
fewer MU and less time (Afghan et al. , 2008; Otto, 2008; Palma et al. , 2008b,a;
Wol et al. , 2008; Yu, 1995). Techniques such as VMAT may negate any delivery
time disadvantages of IMRT. However, it could be assumed the resources required for
planning and quality assurance for VMAT would be similar to that for conventional
IMRT.
In retrospect, the eect of two extra beams and intensity modulation have not been
decoupled from each other. The dosimetric advantages obtained with IMRT may have
been due to a combination of both extra beams and increased modulation. The goal
of this study was to compare current clinical practice with a 'high-quality' IMRT plan;
that is, to make the most of the IMRT plan. Future work could involve replanning all
15 patients using an inverse-planned, single segment per beam method. This would
decouple the eects of the extra two elds and the increased modulation.
2.5. Conclusion
75
2.5 Conclusion
In the report we have shown that for equivalent coverage of the PTV, physically
optimised IMRT can reduce the dose to normal tissue for prostate cancer patients.
From the rectal DVHs it was shown that IMRT reduced the volumes receiving doses
over the whole dose range. Minimal reduction in bladder dose was observed, but this
was not actively pursued as bladder doses were well below tolerance doses. Doses
to femoral heads were reduced due to the use of seven elds (IMRT) as opposed
to ve elds (3DCRT). Reductions in rectal doses were reected in reduced rectal
NTCPs. The IMRT plans resulted in reduced average NTCPs for three sets of model
parameters. These results were statistically signicant. The IMRT plans required on
average 42% more monitor units for delivery.
Chapter 3
Biological optimisation of prostate
IMRT plans
3.1 Introduction
There is some variation in techniques for IMRT optimisation. Variations in target
and organ at risk geometrical delineation, number of elds and their location, and
optimisation parameters exist. No one technique of IMRT optimisation has been
shown to be ideal for prostate IMRT and clinicians are still investigating dierent
ways to improve target dose and reduce OAR dose. Furthermore, the endpoints used
to judge IMRT plans, primarily dose-volume, can vary between physicians. Currently,
there is an interest in using biological end points describing either tumour control or
normal tissue toxicity to rank plans. To this end, optimisation of IMRT plans based
on biological end points is a logical area of investigation. The commercial biological
IMRT optimisation investigated here is based on generalised Equivalent Uniform Dose
(gEUD) (Pinnacle RTPS).
gEUD was introduced as a tool for evaluating and reporting heterogeneous dose
distributions (Niemierko, 1997). gEUD is the dose that if given uniformly will result
76
77
3.1. Introduction
Table 3.1: Conditions and use of the parameter a
Condition
a<0
a=0
a>0
Use
Lower gEUD as a ) - 1. Cold spots are penalised. Suitable for targets
gEUD = geometrical mean. Equal weights given to hot and cold spots
Higher gEUD as a ) 1. Hot spots are penalised. Suitable for organs at risk
in the same biological eect as the heterogeneous dose distribution. The form given
by Niemierko:
! =a
N
X
1
a
di
gEUD =
N
1
i=1
(3.1)
Where N is the number of voxels, di is the dose in voxel i and a is the volume
parameter. The treatment planning system used for this study, Pinnacle V7.6c with
the P3IMRT optimisation toolbox (Philips Radiation Oncology Systems, Fitchburg,
WI, USA), uses a slightly modied form of the gEUD equation to allow voxels to be
only partially included in a ROI:
gEUD =
N
X
i=1
vi dai
!1=a
(3.2)
Where vi is the fractional volume of the ROI occupied by voxel i. The value of
a is equal to n , the volume parameter used in the Lyman-Kutcher-Burman NTCP
model. The value of a aects the gEUD as follows:
The optimisation of IMRT plans using gEUD objectives has been shown to be an
attractive method to obtain a plan not only meeting dose-volume criteria but also
biological constraints (Choi & Deasy, 2002; Wu et al. , 2002; Thieke et al. , 2003;
Schwarz et al. , 2004; Olafsson et al. , 2005; Wu et al. , 2003). Although current
biological models such as gEUD have limitations, as biological systems are extremely
complex, the gEUD parameter may be used as a simple optimisation objective to meet
1
78
3.1. Introduction
dose-volume criteria, without considering resultant NTCP and TCP.
gEUD optimisation of dose to normal tissue can require fewer objectives compared
with physical dose-volume optimisation where multiple objectives are often required
to reduce volumes receiving doses over the whole dose range. The user only has to
provide a maximum or target gEUD value and the parameter a, both of which will
vary with organ. Fewer objectives should lead to a more ecient planning process to
achieve similar or better dose distributions than physically optimised plans.
The Pinnacle RTPS provides optimisation based on gEUD criteria via the biological
optimisation toolbox. Specically, the optimisation function is dened as:
F (gEUD) = (gEUD; gEUD0 )
8
>
>
>
>
>
>
<
gEUD gEUD0
gEUD0
!2
(3.3)
9
>
>
>
>
>
>
=
H (gEUD; gEUD0 )
for max gEUD
(gEUD; gEUD ) = >>
1
for target gEUD >> (3.4)
>
>
>
>
>
>
>
;
: H (gEUD ; gEUD) for min gEUD >
Where H(.) is the Heaviside step function and gEUD is the dose value specied
by the user (Raysearch, 2003; Schwarz et al. , 2004). The objective function acts such
that too high gEUD values are penalised when using a maximum gEUD objective
and too low gEUD values are penalised when using a minimum gEUD objective. The
target gEUD value penalises any deviation from the target gEUD value.
The behaviour of the gEUD value and F(gEUD) as a maximum gEUD objective
function for a given range of a values calculated over three example rectal DVHs is
shown in Figure 3.1. A gEUD value of 50Gy was used in generating these plots.
Figure 3.1a shows example DVHs selected for analysis. DVH1 is a rectal DVH with
low volumes receiving low doses and higher volumes receiving high doses. DVH2 is a
rectal DVH with intermediate volumes receiving high and low doses. DVH3 is a rectal
0
0
0
0
3.1. Introduction
79
Figure 3.1: Behaviour of the gEUD and f(gEUD) functions (a) Example DVHs used
for analysis (b) Change in gEUD as a function of a (c) Optimisation function value as
a function of a and (d) Optimisation function value as a function of gEUD
3.1. Introduction
80
DVH with higher volumes receiving low doses and low volumes receiving high doses.
Figure 3.1b shows the change in calculated gEUD as a function of the parameter a.
gEUD increases with value of a. DVH1 has a lower gEUD with low a value as a low a
value rewards reduction in low doses. As a increases the gEUD increases at a greater
rate than DVH2 and DVH3 as the high volumes at high doses are penalised more with
larger values of a. DVH3 shows a higher gEUD with low a value as the higher volumes
at low doses are penalised. However as a increases the low volumes at high doses are
rewarded with a low gEUD. DVH2 represents an intermediate DVH that is therefore
rewarded at each end with low gEUD. Figure 3.1c shows the result of the optimisation
function F(gEUD) as a function of a. F(gEUD) is highest for DVH1 at high a values
as the high volumes at high doses is getting penalised. It then decreases the fastest to
its minimum as a decreases as the low volumes at low doses is rewarded. DVH3 has
the lowest F(gEUD) value at high a values due to the low volume at high doses. Figure
3.1d F(gEUD) vs gEUD showing that for the three example DVHs the optimisation
function reduces the gEUD at the same rate until it is the maximum gEUD (50Gy in
this case) or lower.
The aim of this chapter is to demonstrate the utility of maximum gEUD optimisation objectives for the rectum in prostate IMRT plans. This is achieved by creating
prostate IMRT plans for sixteen patients using a maximum gEUD objective on the
rectum and physical dose objectives on the target and other organs at risk. The parameter a is varied to nd the value that has the largest eect on minimising rectal
dose.
81
3.2. Methods and materials
3.2 Methods and materials
3.2.1 Treatment planning
The Pinnacle Radiotherapy Treatment Planning System (RTPS) (Philips Radiation
Oncology System, Fitchburg, USA) with the biological toolkit was used for all plans
(Raysearch, 2003). The same group of 16 prostate patients as used in Chapter 2 were
used for analysis. Three, seven-eld IMRT plans were generated for each patient. All
plans had elds with xed gantry angles (IEC Convention) 120, 80, 40, 0, 320,
280 and 240. The plans were optimised using biological objectives on the rectum
and physical dose-volume objectives on all other ROIs. All plans were optimised using
Direct Parameter Machine optimisation (DMPO). A maximum of 70 segments were
used for each plan and 50-75 iterations were run for each plan.
As the eect of biological optimisation parameters was being investigated, the PTV
DVH coverage was kept constant for all plans. For the rectal objectives, the parameter
a was varied, with values of 3, 4.5 and 9 used. A large variation in the value of the
parameter a for the rectum exists in the literature (Rancati et al. , 2004; Tucker et al.
, 2004a; Sohn et al. , 2007). While the value of a should depend on the end point
being investigated, with lower values of a arising from parallel organ type toxicities
and high values of a arising from serial organ type toxicities, values ranging from
0-32.3 (95% CI) have been reported in the literature for the same endpoint (Tucker
et al. , 2004a). The parameter a however can be used as purely an optimisation tool,
in that it can be varied to obtain the best dose distribution using trial and error.
The values of the parameter a were chosen in this fashion; a range of values were
selected for analysis. The maximum gEUD dose value, equal to the value of EUD in
the optimisation function (Equation 3.4), was reduced to as low as possible without
compromising PTV coverage. The maximum gEUD values were between 35Gy and
0
3.2. Methods and materials
82
Table 3.2: Optimisation parameters used in biological IMRT plans
ROI
Objective Type Dose (Gy)a % Volume Weight
a
min dose
74.1
constraint
PTV
max dose
81.0
constraint
min DVH
76
95
100
Rectum
max gEUD
ALAP
40
3, 4.5, 9
Bladder
max DVH
50
ALAP
30
Femoral Heads max dose
50
30
Ring
max DVH
74.1
5
25
60Gy for all plans. The optimisation objectives are given in Table 3.2.
3.2.2 Plan analysis
All plans were imported into the Computational Environment for Radiotherapy Research (Deasy et al. , 2003) (CERR, University of Washington in St. Louis, USA).
From the CERR platform the target and organ at risk DVHs, rectal gEUDs and rectal
NTCPs were calculated for comparison.
Rectal gEUDs were calculated for all patients using the values of a used for optimisation. This was done to test the ability of the maximum gEUD optimisation to
act to reduce EUD with specically dened a values, as set in our optimisation.
The NTCP calculation was performed using the Lyman-Kutcher-Burman (LKB)
NTCP model with the EUD reduction scheme (Lyman, 1985; Kutcher & Burman,
1989). Three parameter sets were used, so as to minimise any eects of variations
of model parameters. The three parameter sets were chosen from the literature, all
representing grade 2 or worse rectal bleeding. The parameter values are given in Table
3.3.
3.3. Results
83
Table 3.3: NTCP calculation parameters
Parameter Set n
m D50
Reference
NTCP1
1.03 0.16 55.9 Tucker et al. (Tucker et al. , 2004a)
NTCP2
0.24 0.14 75.7 Rancati et al. (Rancati et al. , 2004)
NTCP3 0.084 0.108 78.4 Sohn et al. (Sohn et al. , 2007)
3.3 Results
3.3.1 Dose-volume histograms
Average cumulative DVHs are given in Figure 3.2. Figure 3.2a shows the average
cumulative DVHs for the PTV and the rectum. Equivalent PTV coverage is seen.
The average rectal DVHs change with the optimisation parameter a ; as a increases,
the volume receiving mid-low doses increases and the volume receiving high doses
decreases. Figure 3.2b shows the average cumulative DVHs for the bladder and the
femoral heads. As the value of a changes, no change in the bladder DVH is observed.
However, an indirect consequence of increased a values is that the dose to the femoral
heads decreases.
The shape of the rectal DVHs is dependent on the value of a used in the optimisation. As a high value of a equates to more serial tissue architecture, a greater
emphasis is placed on reducing the high dose region. This leads to decreased volume
receiving high doses. Minimal emphasis is placed on reducing mid-low doses; as a
result, higher volumes receive these doses. A low value of a represents more parallel
tissue architecture therefore a greater weight is placed on reducing the mean dose to
the organ. Thus the emphasis is placed on reducing the mid-low doses.
3.3. Results
84
Figure 3.2: Average cumulative DVHs over all 16 patients for a) PTV and rectum and
b) bladder and femoral heads
3.3. Results
85
3.3.2 gEUD comparison
The gEUD was calculated for the three plans using the same values of a used in
optimisation, that is, a =3, 4.5 and 9. The average calculated gEUDs for the three
plans, as calculated with a values of 3, 4.5 and 9 are given in Figure 3.3.
When the gEUD is calculated with a =3, plan gEUD (a =3) had the lowest average
gEUD (39.64Gy vs 45.79Gy and 42.92Gy for gEUD (a =4.5) (p<0.001) and gEUD
(a =9) (p<0.001) respectively). When the gEUD is calculated with a =4.5, plan gEUD
(a =4.5) had the lowest gEUD (51.47Gy vs 51.72Gy and 52.25Gy for gEUD (a =3)
(p>0.2) and gEUD (a =9) (0.02<p<0.05) respectively). For the gEUD calculated
with a =9, plan gEUD (a =9) resulted in the lowest gEUD (60.10Gy vs 61.16Gy and
60.45Gy for gEUD (a =3) (0.005<p<0.01) and gEUD (a =4.5) (p>0.2) respectively.
3.3.3 NTCP comparison
The rectal NTCPs were compared for the three plans using three sets of LKB model
parameters (Table 3.3). These three sets of parameters all represent the same toxicity
endpoint (grade 2 rectal bleeding), but cover a wide range of values. It is well
understood why such a wide variability of model parameters exist in the literature
but it is suspected that variations in treatment technique and patient genetics are
the cause. The rectal NTCPs for all 16 patients for the three sets of parameters are
given in Figure 3.4. The average NTCPs over all patients are given in Table 3.4. For
parameter set NTCP1, the lowest average NTCP was achieved with plan gEUD (a =3)
(1.29% vs 1.60% and 3.59% for gEUD (a =4.5) (0.02 < p < 0.05) and gEUD (a =9)
(p < 0.001) respectively). For parameter set NTCP2, the lowest average NTCP was
achieved with plan gEUD (a =4.5) (1.47% vs 1.66% and 1.77% for gEUD (a =3) (P >
0.2) and gEUD (a =9) (0.005 < P < 0.01) respectively). For parameter set NTCP3, the
lowest average NTCP was achieved with plan gEUD (a =9) (4.02% vs 5.63% and 4.58%
86
3.4. Discussion
Table 3.4: Average rectal NTCPs over all 16 patients
Plan
Parameter Set
NTCP1 (%) NTCP2 (%) NTCP3 (%)
gEUD (a =3) 1.29
1.66
5.63
gEUD (a =4.5) 1.60
1.47
4.58
gEUD (a =9) 3.59
1.77
4.02
Table 3.5: Average MUs
Plan
Average MUs Range
gEUD (a =3)
534
452-649
gEUD (a =4.5)
516
436-628
gEUD (a =9)
468
412-577
for gEUD (a =3) (P < 0.001) and gEUD (a =4.5) (0.005 < P < 0.01) respectively).
3.3.4 Delivery eciency
The monitor units required for delivery were recorded for each plan. The average MUs
for a 200cGy fraction are given in Table 3.5. As the value of a used in optimisation
increased, the required monitor units decreased. A possible reason for this is that as
a increases, the optimiser is only trying to reduce the volumes receiving high doses.
Limited gains can be made in this region as it corresponds to the anterior rectum
that overlaps with the PTV. This results in a reduced amount of modulation therefore
fewer MUs are required to achieve adequate dose.
3.4 Discussion
We have undertaken a study to determine the optimal value of a used in gEUD objectives for the rectum in prostate IMRT. The purpose of this study was to determine
what eect changing the value of a had on the rectal DVHs and the calculated gEUDs
3.4. Discussion
87
Figure 3.3: Average calculated gEUDs over all 16 patients for the three values of a
used in planning
and NTCPs. This problem was approached purely in terms of optimal parameters for
using gEUD optimisation as a planning tool, rather than optimising to a denitive
biological end point. Due to the large range of values of a in the literature, optimising
with a specic set of biological endpoint values taken from the literature could be
problematic and not result in the optimal plan.
The calculated gEUDs demonstrate that the only statistically signicant reduction
in desired gEUD over the plans optimised with the other two values of a occurs when
optimising with a =3. This value of a represents more parallel tissue architecture
and as such the mid-low doses are reduced when optimising with this value. When
compared with the rectal DVHs it can be said that when optimising with a lower value
of a, rectal dose reductions are seen over a greater dose range. In terms of anatomy
this equates to minimising the dose to the posterior and middle portions of the rectum
by increasing the dose gradient from the PTV.
A higher value of a equates to a more serial organ architecture and as such opti-
3.4. Discussion
88
Figure 3.4: Rectal NTCPs for all 16 patients calculated with a) NTCP1 b) NTCP2
and c) NTCP3
3.5. Conclusion
89
mising with a higher value of a will target aim to reduce volumes receiving high doses.
This approach has its limitations. The PTV encompasses a section of the anterior
rectum. Since the PTV must receive the target dose, the portion of the rectum overlapping with the PTV will always get the target dose. As a result, minimal gains can
be made when optimising with a high value of a.
Optimising with a =3 resulted in a slight increase in volumes receiving high doses
(>65Gy). This result is similar to that obtained by Schwarz et al (Schwarz et al.
, 2004). Therefore it is suggested that the higher volumes receiving low doses can
be alleviated by coupling the maximum gEUD objective with a maximum dose/DVH
objective on the rectum. The maximum gEUD objective minimises the volumes receiving mid-low doses and the maximum dose/DVH objective limits the high dose
region. The maximum dose/DVH objective would primarily reduce the dose to the
rectum in the overlap region between the PTV and the rectum and as such should be
set no less than the minimum dose required for the PTV. This approach was taken in
Chapter 4, where a maximum gEUD objective with a =3 was used in conjunction with
a maximum dose objective of 76Gy on the rectum for a 78Gy target prescription dose.
Another approach would be to use two maxiumum gEUD objectives on the rectum one with a low value of a and one with a high value of a.
3.5 Conclusion
Biological optimisation with the maximum gEUD parameter was shown to be a feasible
and eective method of reducing rectal doses in prostate IMRT. The maximum gEUD
function acts to reduce the volumes receiving doses over a large range, rather than
single DVH points as in DVH optimisation. Using an a value of 3 resulted in the
optimal reduction in gEUD in the mid-low dose range however this resulted in a slight
increase in volumes receiving high doses. It is suggested that coupling a maximum
3.5. Conclusion
90
gEUD objective with a maximum dose or maximum DVH objective for the rectum
provides optimal rectal dose reduction.
Chapter 4
Comparison of prostate IMRT and
VMAT biologically optimised
treatment plans
4.1 Introduction
Intensity Modulated Radiation Therapy (IMRT) has been shown to be the preferred
delivery method for prostate radiotherapy (Zelefsky et al. , 2000, 2001; Kupelian et al.
, 2002b,a; Sanguineti et al. , 2006; Namiki et al. , 2006; Veldeman et al. , 2008). A new
method of delivery has recently become available, Volumetric Modulated Arc Therapy
(VMAT) (Yu, 1995; Otto, 2008; Bzdusek et al. , 2009). VMAT is the delivery of IMRT
while the linac is in rotation. This is essentially an open aperture IMRT arc technique.
Parameters that can be varied are dose rate, gantry speed and number of arcs. One
perceived benet of VMAT is the increase in delivery eciency.
Part
of this chapter has been submitted for publication in Radiotherapy and Oncology:
Hardcastle, N., Tom
e, W. A., Foo, K., Miller, A., Carolan, M. and Metcalfe, P. E. (2009).
Comparison of prostate IMRT and VMAT biologically optimised treatment plans. Medical Dosimetry (in submission)
91
4.1. Introduction
92
Figure 4.1: Example dose distributions for IMRT (left) and VMAT. Dose scale on the
right is in Gy.
Recent planning studies have compared VMAT with conventional delivery techniques such as IMRT, 3D conformal (3DCRT) and tomotherapy for prostate radiotherapy (Palma et al. , 2008a,b; Wol et al. , 2008). Afghan et al. (2008) compared
VMAT with step and shoot IMRT for ve prostate patients and found delivery times
decreased by up to 54% for equivalent target coverage and normal tissue sparing.
Palma et al. (2008a) compared constant and variable dose rate VMAT with IMRT
and 3DCRT plans for ten prostate patients. Improved rectal, bladder and femoral
head sparing was observed with the VMAT plans over the IMRT and 3DCRT plans.
Fewer monitor units were required with the VMAT plans than for the IMRT plans,
but not for the 3DCRT plans. Kjaer-Kristoersen et al. (2008) achieved equal or
better normal tissue sparing for prostate treatments with VMAT over conventional
IMRT, however decreased target dose homogeneity was observed. This was due to a
decrease in the level of modulation achievable with VMAT, as discussed in Chapter 1.
The aims of recent studies in the literature have been steered towards the delivery
eciency of VMAT and whether equivalent target and organ-at-risk doses can be
delivered while maintaining delivery eciency. This study diers in that it aims to
compare biologically optimised VMAT plans with IMRT plans using physical dose and
biological endpoints as well as delivery eciency as the comparison metrics.
4.2. Methods and materials
93
4.2 Methods and materials
4.2.1 Treatment planning
Ten prostate radiotherapy patients were selected for analysis. These patients were
part of the 16 patients used in Chapters 2-3 however not all patients were used due
to le transfer issues. The target was the prostate not including the seminal vesicles.
A uniform 7mm CTV-PTV margin was used. The prescription dose was 78Gy in 34
fractions (Pollack et al. , 2002; Kuban et al. , 2008). The rectal volume was dened
using full volume and the length from anal canal up to the anterior curve of rectum
into sigmoid colon with a maximum distance of 11cm.
The Pinnacle RTPS (Philips Radiation Oncology Systems, Fitchburg, USA) with
the biological optimisation tool kit was used (Raysearch, 2003). All plans were created
for a Varian 21EX (Varian Medical Systems, Palo Alto, USA) with 120 leaf multi-leaf
collimator. For each patient two treatment plans were created. The rst was a seven
eld IMRT plan from xed gantry angles 120, 80, 40, 0, 320, 280 and 240 (IEC,
1996). The second plan was created with an alpha version of the Pinnacle SmartArc
planning tool. A single arc utilising the full 360 was employed. Dose rate modulation
was used and a maximum delivery time of 120s was set. The same number of iterations
(50) was run for the IMRT and VMAT plans.
For both the IMRT and VMAT plans the optimisation objectives given in Table 4.1
were used. Biological optimisation with maximum generalized EUD objective was used
on the rectum in all plans. As discussed in Chapter 3, a maximum gEUD objective
with a =3 works to reduce the volumes receiving mid-low doses. The maximum gEUD
objective was coupled with a maximum dose objective to ensure the high dose region
was minimised. A maximum DVH objective was used for the bladder, with the volume
receiving 50Gy set to as low as possible (ALAP) without compromising PTV coverage.
4.3. Plan analysis
94
Table 4.1: Optimisation objectives for all IMRT and VMAT plans
ROI
Type Dose (Gy) Volume Weight a
PTV
Min Dose
78
100 Max Dose
81
100 Rectum
Max gEUD 30 - 40
30 3
Max Dose
76
30 Bladder
Max DVH
50
ALAP
3 Femoral Heads Max Dose
50
2.5 A maximum dose objective of 50Gy was set for the femoral heads.
4.3 Plan analysis
All plans were imported into the Computational Environment for Radiotherapy Research (CERR) (University of Washington in St. Louis, USA) (Deasy et al. , 2003).
Cumulative dose volume histograms (DVHs) were compared for the PTV, rectum,
bladder and femoral heads. Comparison between the plans was also done using a biological end point parameter, Normal Tissue Complication Probability (NTCP), calculated for the rectum. The Lyman-Kutcher-Burman (LKB) NTCP model was chosen
(Lyman, 1985; Kutcher & Burman, 1989). To minimize the any impact of LKB model
parameters, the same three sets of LKB model parameters were chosen from the literature (as for Chapter 3), all representing Grade 2 rectal toxicity. The NTCP
parameters are given in Table 4.1. Although the LKB model parameters by Tucker
et al. (2004b) have recently been updated (Tucker et al. , 2007), the NTCP1 values in
Table 2 were used so as to represent NTCPs over a range of model parameters. The
NTCPs were calculated using the CERR tool kit.
Delivery eciency was assessed by comparing the number of monitor units (MUs)
and delivery time. The average number of MUs for each IMRT and VMAT plan was
compared. The delivery time for all VMAT plans was taken as 120s, the maximum
4.4. Results
95
Table 4.2: NTCP calculation parameters
Parameter Set n
m D50 (Gy)
Reference
NTCP1
1.03 0.16 55.9 Tucker et al. (2004b)
NTCP2
0.24 0.14 75.7 Rancati et al. (2004)
NTCP3 0.084 0.108 78.4
Sohn et al. (2007)
delivery time set in optimisation. The delivery time for the IMRT plans was approximated by taking the average delivery time for each fraction for ten patients at our
institution that have undergone seven eld IMRT. The 'beam on' time was dened as
the time from the start of the rst beam to the end of the last beam. The Wilcoxon
matched-pair signed-rank test was used to compare the DVHs, NTCP and MU results
between IMRT and VMAT plans with a statistical signicance threshold of p < 0.05.
4.4 Results
4.4.1 Dose-volume histograms
Resultant IMRT and VMAT dose distributions from one patient are given in Figure
4.1. The individual patient rectal and PTV DVHs are given in Figure 4.2. The average
cumulative DVHs for the PTV and rectum are given in Figure 4.3a. For equivalent
PTV coverage, VMAT plans on average result in lower rectal volumes for doses <
50Gy. VMAT and IMRT plans have similar volumes receiving doses > 50Gy.The
average cumulative DVHs for the femoral heads and bladder are given in Figure 4.3b.
VMAT and IMRT resulted in similar average DVHs. Both VMAT and IMRT plans
resulted in similar maximum femoral head dose however VMAT plans resulted in higher
volumes receiving doses over the whole range of 0- 50Gy. Table 4.3 shows average DVH
parameters for the IMRT and VMAT plans for the Rectum. Statistically signicant
reduction in the V25Gy parameter and increase in the V70Gy parameter was seen for
96
4.4. Results
Figure 4.2: PTV and rectal DVHs for all 10 patients
4.4. Results
97
Table 4.3: Summary of average DVH parameters over the ten patients
ROI Parameter IMRT VMAT
p-value
Rectum V25Gy 51.7 43.5
< 0.01
V50Gy 20.8 20.3 not signicant
V60Gy 15.1 15.4 not signicant
V70Gy 9.6 10.0
< 0.01
V75Gy 5.9
5.9 not signicant
Table 4.4: Summary of average NTCPs for IMRT and VMAT plans
Parameter IMRT VMAT
p-value
NTCP1 (%) 0.91 0.55
< 0.01
NTCP2 (%) 0.81 0.77 not signicant
NTCP3 (%) 3.57 3.58 not signicant
the Rectum with VMAT.
4.4.2 NTCP comparisons
Figure 4.4 shows the calculated NTCP for all ten patients in the study. Figure 4.4a,
Figure 4.4b and Figure 4.4c show the NTCPs for parameter set NTCP1, NTCP2 and
NTCP3 respectively. For parameter set NTCP1, VMAT results in a lower NTCP
for all ten patients. Parameter set NTCP1 has an n value that corresponds to more
parallel tissue architecture. As a result, the gains in rectal DVH with VMAT at the
low-mid dose ranges are reected in the NTCP. For Parameter set NTCP2, VMAT
has a higher rectal NTCP for one patient. There is minimal gain, if at all, for nine out
of the ten patients. Parameter set NTCP2 has an n value reecting more serial organ
architecture. As VMAT and IMRT rectal DVHs are very similar at high range doses
then no gain in NTCP is made with VMAT or IMRT. This is further accentuated with
parameter set NTCP3, which has an even lower n value.
4.4. Results
98
Figure 4.3: Average cumulative DVHs of a) PTV and rectum and b) bladder and
femoral heads for IMRT and VMAT plans.
4.4. Results
99
Figure 4.4: NTCPs for IMRT and VMAT plans for all 10 patients (a) NTCP1 (b)
NTCP2 and (c) NTCP3
100
4.5. Discussion
Table 4.5: Delivery eciency: Average required MUs and delivery time for IMRT and
VMAT plans
Plan Average MU SD p-value Average delivery time (min:sec) SD
IMRT
526 63
7:31 2:06
<
0.01
VMAT
417 44
2:00 0
4.4.3 Delivery eciency
Figure 4.5 shows the required monitor units for delivery of each plan. In our clinic,
1MU is equal to 1cGy at Dmax (1.5cm) in water for a 10x10cm beam with 6MV
photons. The average MU and delivery time data is given in Table 4.5. Averaged over
all ten patients, VMAT required 18.6% fewer monitor units than IMRT (521MU vs
424MU respectively, p < 0.01) for delivery of a 2Gy fraction. The average 'beam on'
time for a seven eld IMRT plan at our institution was 7min 31sec, compared to the
maximum delivery time for VMAT of 2min.
2
4.5 Discussion
An average reduction in the rectal V25Gy values over the ten patients of 8.20% was
observed. This reduction in V25Gy came at the expense of a minor increase in V70Gy
of 0.44%, averaged over the ten patients. It can be argued that the gains made with
the V25Gy parameter are greater than the increase in rectal toxicity probability with
the higher V70Gy. This is seen in the NTCP values for parameter sets NTCP2 and
NTCP3, where no statistically signicant increases in rectal NTCP were observed.
The three parameter sets used allow some estimate of the range of NTCP values
that might be experienced for grade 2 rectal toxicity, given the variability in the
published model parameters. The value of n is 1.03, 0.24 and 0.084 for NTCP1,
NTCP2 and NTCP3 respectively. Consequently, each parameter set penalises the
rectal DVHs according to the rectum being more parallel (low n ) or serial (high n ). It
101
4.5. Discussion
Figure 4.5: Total MU for all ten patients
should also be acknowledged that clinically, there may be various toxicity end points to
be considered. Each of these end points will be characterised by its own set of NTCP
model parameters. As a result, the clinical selection of a plan based on NTCP could
involve the assessment of a range of NTCP values related to various other toxicity end
points, as suggested by Rancati et al. (2004). These NTCP values may span a similar
range to the uncertainty in the NTCP for any single endpoint such as the range of
grade 2 rectal toxicities which we have calculated here.
The high dose region of the rectum is included in the PTV, so the only reduction
in NTCP due to this dose can be made by reducing the CTV-PTV margin. It is
very dicult to achieve reduction of the dose to this region with technique changes.
Therefore, as the NTCP parameter set becomes more weighted towards the small
volume of the rectum receiving high doses, minimal change in NTCP is seen when
technique is changed.
The MU reduction observed in this study is consistent with results obtained by
4.6. Conclusion
102
Afghan et al. (2008) and Palma et al. (2008a) for prostate irradiation. The results
in this study however show a smaller reduction in required monitor units; any number
of factors can lead variations in delivery eciency, as discussed by Ost et al. (2009)
and Palma et al. (2009). The reduction in monitor units suggests that less beam
modulation is occuring. This is most probably a consequence of the target shape, in
that it is a simple spherical target for which more open segments (less modulation) is
adequate for target coverage.
The reduction in the required monitor units for delivery is overshadowed by the
large gains in 'beam on' time. At the Illawarra Cancer Care Centre, a seven eld IMRT
plan takes on average 7minutes 31seconds of 'beam-on' time to deliver, approximately
3.75 times longer than would be achievable with VMAT. This is a very important gain
in delivery eciency which may signicantly increase patient throughput. Additionally, institutions not yet employing IMRT for localised prostate radiotherapy due to
increases in delivery time over 3DCRT, may view VMAT as an attractive modality to
improve plan quality whilst maintaining delivery eciency.
4.6 Conclusion
This study has compared IMRT and VMAT plans for ten prostate patients. We have
shown that for equivalent target coverage, reduction in rectal doses, specically the
V25Gy parameter, can be achieved using a single-arc VMAT for prostate radiotherapy.
This reduction in rectal dose was achieved with an average of 18.6% fewer monitor
units per fraction. A statistically signicant average reduction in rectal NTCP was
achieved for one rectal NTCP parameter set that penalises volumes receiving low-mid
doses, when employing a maximum gEUD constraint with a =3. In general, provided
the isodose maps and DVHs show similar target coverage, the rectal DVH reduction
observed with VMAT plans result in a superior plan.
Chapter 5
Optimisation of prostate IMRT
plans based on a theoretical 'goal'
dose
5.1 Introduction
The optimisation of IMRT treatments is based on prior knowledge of achievable dose
distributions. In general, objectives are applied to target volumes and organs at risk
(OARs) that result in a dose distribution in which the target volumes receive a prescribed dose and the OARs receive as little dose as possible. Limitations exist when
target volumes and OARs intersect; priority is usually placed on target volumes to
maximise the eectiveness of the treatment.
The planner takes into consideration the prescription dose and any intersections of
OARs and target volumes when setting objectives. Take the case of prostate cancer
radiotherapy. There is the prostate gland, commonly set as the clinical target volume
(CTV). A margin is then applied to the CTV to result in the planning target volume
(PTV). Immediately adjacent to the CTV is the rectum, so when the CTV-PTV
103
5.2. Method
104
margin is applied the PTV encompasses some volume of the rectum. When setting
IMRT objectives the planner understands that some part of the rectum then receives
at least the minimum dose to be received by the PTV.
In Chapter 3, it was shown that the greatest dosimetric advantage when using the
maximum generalised Equivalent Uniform Dose (gEUD) IMRT optimisation function
was achieved using a=3 as the exponent in the gEUD equation. Rectal dose reduction
using the maximum gEUD optimisation function requires two inputs - the exponent, a,
and the target maximum gEUD. In Chapter 3, the target maximum gEUD was found
using an iterative process for each individual patient whereby the maximum gEUD
was reduced as low as possible without impacting on the PTV dose. This process
requires constant monitoring and updating of the values. It is proposed that if there
is prior knowledge of the 'goal' dose distribution for a given patient's anatomy, the
'goal' DVH or gEUD can be used as a goal for the optimisation. If the 'goal' gEUD
is obtained, both parameters required for the objective function would then be known
prior to optimisation, and the iterative process required to reduce the maximum gEUD
for the rectum would not be required.
This study investigates the use of an articially created 'goal' dose distribution
created outside of the planning system for prostate cancer. IMRT optimisation parameters are then guided by this 'goal' dose distribution. The planned dose distributions
were compared with the goal dose distributions and previously planned distributions.
5.2 Method
The 10 prostate patients used in Chapter 4 were chosen to demonstrate the algorithm.
In brief, the algorithm:
Obtain CT scan
5.2. Method
105
Delineate target and OARs
Generate optimal dose distribution based on prescription dose and anatomy
Calculate EUD and/or NTCP from goal DVH
Set IMRT optimisation parameters based on physical dose-volume points or biological parameters
Run optimisation to achieve dose distribution as close as possible to the goal
distribution
5.2.1 Contouring
The whole prostate without seminal vesicles was contoured as the CTV. A 7mm margin
was applied to the CTV to obtain the PTV. Rectal volume was dened using full
volume and the length from anal canal up to the anterior curve of rectum into sigmoid
colon with a maximum distance of 11cm. The femoral heads, bladder and normal
tissue (external contour - CTV) were also contoured. A 7mm margin was also applied
to the rectum to obtain the planning rectal volume (PRV) OAR in anticipation of
rectal movement and change in volume. A 1cm margin was applied to the PTV to
obtain the penumbra ROI.
In addition to these contours, four extra optimisation contours were created - 95%
zone, 100% zone, penumbral zone and scatter zone. The derivation of the optimisation
contours is given in Table 5.1. The contours are shown in Figure 5.1. The contours were
generated using tools in the Computational Environment for Radiotherapy Research
(CERR, University of Washington in St. Louis, USA) Deasy et al. (2003).
5.2. Method
106
Table 5.1: Derivation of optimisation contours
Zone
Derivation
Dose
100% zone
PTV-PRV
78Gy
95% zone
PRV \ PTV
74.1Gy
penumbral zone PRV \ Penumbra Gradient from 95% dose ! scatter dose
scatter zone PRV - Penumbra
10Gy
Figure 5.1: Contours used for IMRT optimisation. Red = 100% zone, Light Red =
95% zone, Orange = penumbral zone, green = scatter zone and purple = rectum.
107
5.2. Method
5.2.2 Goal dose distribution
An optimal dose distribution was created so that a goal DVH could be used in the
optimisation process. The CT data and contours were imported into the CERR platform. A goal dose matrix was created manually that represents what could possibly
be delivered using photon beams. That is, target dose with penumbral and scattered
dose regions were created. Doses were only created for dosimetric regions of interest
(the PTV and the rectum).
Every voxel in the 100% zone was set to 78Gy. Each voxel in the 95% zone was
set to 74.1Gy. Each voxel in the penumbral zone was set to a gradient from 95% to
10% based on the distance outwards from the PTV. This gradient was a conservative
estimation based on a 10x10cm eld at 10cm depth in water. The dose in the scatter
zone was set to 10Gy. The MATLAB programming environment was used for creation
of the goal dose distribution. The code is presented in Chapter 10.9.5, Section A.
Briey, the code interfaces with the treatment plan and contours which are saved in
the CERR format. Three-dimensional masks of 1s and 0s were generated based on
the four described contours (1 inside the contour, 0 everywhere else). A mask for each
relevant contour was created and each mask was multiplied by its respective dose to
result in individual dose matrices for each of the four contours. These dose matrices
were then summed and the resultant dose distribution saved back into the treatment
plan in the CERR format for analysis. The resultant dose distribution is given in
Figure 5.2.
The 'goal' dose distribution was created for each of the 10 prostate patients used
in Chapter 4. The resultant DVHs were then obtained for the rectum and PTV
and compared with the DVHs obtained using biologically optimised seven eld IMRT
plans optimised with a=3, in Chapter 3. The DVHs for the rectum and PTV for all
10 patients is given in Figure 5.3. The discrete dose levels selected for the penumbra
2
108
5.2. Method
Figure 5.2: Goal dose distribution created in MATLAB
(10 levels over 1cm) plus the 100%, 95% and scatter zones result in the discrete steps
in the cumulative DVH curve.
5.2.3 IMRT optimisation
A seven-eld IMRT plan was created using the optimisation parameters given in Table
5.3. A maximum of 70 segments was set. The number of iterations was set to 25. For
the rectum, a maximum gEUD objective was set using a=3. The maximum gEUD
value was taken as the 'goal' gEUD for each patient as presented in Table 5.2. One
set of iterations was performed. The resultant dose distributions were exported into
CERR for analysis.
5.2. Method
109
Figure 5.3: The 'goal' DVH for all 10 patients compared with the seven eld IMRT
DVHs obtained in Chapter 3 with a=3. The seven eld IMRT plan obtained by
optimising based on the 'goal' DVH is also shown ('planned goal').
110
5.2. Method
Table 5.2: Calculated rectal gEUDs from goal DVHs using a=3
Patient gEUD (Gy)
1
30.6
2
29.7
3
36.3
4
37.7
5
38.8
6
38.3
7
48.0
8
47.0
9
38.9
10
40.2
ROI
PTV
PTV
PTV
rectum
rectum
bladder
fem heads
Table 5.3: IMRT optimisation parameters
Type
Dose (Gy)
Volume Weight
min dose
74.1
- constraint
max dose
81
- constraint
min DVH
76
95
40
max EUD [as per Table 5.2] 30
max dose
76
30
max DVH
50
ALAP
30
max dose
50
225
a
3
-
5.3. Results
111
5.3 Results
The resultant dose distributions are shown in Figure 5.4. The resultant PTV and
rectal DVHs are shown in Figure 5.3. From Figure 5.3, it can be seen that neither
of the real rectal DVHs meet the 'goal' DVH. It is also seen in Figure 5.3 that the
'planned goal' rectal doses are either equal or lower to the 'EUD (a=3)' rectal doses,
for approximately equivalent target coverage.
The 'EUD (a=3)' plans were obtained using 50-75 iterations. That is, 2-3 optimisation processes were required to reduce the rectal doses to the presented values. Using
a previously derived 'goal' EUD value as an optimisation goal allowed equal or better
sparing to be achieved in 25 iterations (or one optimisation process). Therefore, for
all ten patients the optimisation time was reduced to half or a third of that if no prior
knowledge of the optimal dose distribution exists. The reductions in optimisation
time can be signicant, particularly when employing full convolution/superposition
dose calculations at intermediate and nal stages of the optimisation process.
5.4 Discussion
This chapter details a method of optimising prostate IMRT plans using prior knowledge
of the optimal dose distribution. The optimal dose distribution is highly dependent
on the anatomy of the patient and on the delivery technique. A larger rectum that
spreads away from the target will allow for an improved relative volume DVH as a
larger proportion of the rectal volume can be spared. It is not inherently obvious
how much sparing can be achieved therefore prior knowledge is required to best take
advantage of any anatomically favourable cases.
It can be noted that the presented 'goal' rectal volumes receiving doses from 1040Gy are consistently signicantly less than that achieved in actual treatment plan
112
5.4. Discussion
Figure 5.4: Resultant IMRT dose distribution
optimisation. This suggests that the dose gradient away from the target is too steep
than can be achieved using the seven eld IMRT approach in this study. It is known
that an increase in beam angles allows greater reduction of the mid-low rectal doses
and this was shown in Chapter 4. A more rigorous method of obtaining the penumbral
doses for multiple beams could be employed. It is suggested that a general penumbral
shape could be derived for each nite number of beams used. That is, the penumbral
width would decrease as the number of beams used for delivery increases. The penumbral width would reach a minimum when arc therapy such as tomotherapy or VMAT
are used.
A further improvement could be to use a larger number of discrete dose levels in
the penumbra and 'scatter' region. This would be most important for the dose range
<40Gy, where the largest discrepancies between optimal and deliverable doses exists.
The presented study calculates the gEUD of the 'goal' dose distribution for use
5.5. Conclusion
113
in biological objective based IMRT. The method proposed here can also be employed
using physical dose-volume objectives based on the goal DVH.
The method has been demonstrated for the more simpler case of prostate cancer.
This method has not been tested on more complex IMRT cases such as that seen in
pelvic node or head and neck IMRT. The more complex geometries observed in these
regions may inhibit the ability of optimiser to meet a 'goal' dose distribution. This
warrants further investigation.
5.5 Conclusion
A method of reducing optimisation time by using prior knowledge of the optimal dose
distribution for prostate cancer IMRT has been presented. The method generated an
optimal deliverable photon dose distribution based on the anatomy of the patient. A
'goal' DVH was then calculated for this optimal dose distribution from which the gEUD
was calculated. The gEUD value was used as the optimisation goal for a maximum
gEUD IMRT objective function. The optimisation algorithm was able to return an
equal or superior dose distribution in fewer iterations than what was achieved without
prior dose distribution knowledge.
Chapter 6
Rectal balloon dosimetry in
prostate radiotherapy
6.1 Introduction
6.1.1 Dose escalation and rectal balloons
Local control and disease free survival rates are known to increase with increased dose
in prostate radiotherapy (Hanks et al. , 1998; Zelefsky et al. , 1998; Pollack et al.
, 2000). The ability to increase dose however is limited by toxicity to surrounding
tissues, mainly the rectum (Marzi et al. , 2007; Vavassori et al. , 2007). Rectal toxicity
is directly related to the dose received by the rectal wall (Storey et al. , 2000; Tucker
et al. , 2004b). Reduction in planning target volume (PTV) margins allows reduction
of normal tissue toxicity by reducing the dose delivered to the rectum, which in turn
allows an increase in prescribed dose. The ability to reduce the PTV margins in
prostate radiotherapy requires management of target motion and the ability to reduce
Part
of this chapter has been published in Radiotherapy and Oncology:
Hardcastle, N., Metcalfe, P. E., Rosenfeld A. B. and Tom
e,
W.
A.,
2009,
Endo-rectal balloon cavity dosimetry in a phantom: Performance under IMRT and helical tomotherapy beams.
Radiotherapy and Oncology, volume 92, issue 1, pages 48-56
114
6.1. Introduction
115
the volume of the rectal wall receiving high doses.
Rectal balloons are employed in prostate radiotherapy by a number of institutions
as a means of immobilizing the prostate and reducing the volume of the rectal wall in
the high dose region (Teh et al. , 2001; McGary et al. , 2002; Wachter et al. , 2002;
Patel et al. , 2003; van Lin et al. , 2007).Target immobilization is achieved through the
balloon forcing the prostate against the pubic symphysis. The volume of the rectal
wall receiving high doses is reduced by forcing the posterior rectal wall away from the
target and reducing the rectal wall thickness by expanding the rectal volume. The use
of rectal balloons during treatment delivery has been shown to decrease the delivered
rectal dose (Patel et al. , 2003). Decreased rectal toxicity when using a rectal balloon
has also been reported (Sanghani et al. , 2004; D'Amico et al. , 2006; van Lin et al. ,
2007).
6.1.2 The air cavity eect
Rectal balloons are commonly lled with air in photon radiotherapy, with volumes
of up to 100cm used. The volume of air will perturb the dose distribution in the
surrounding tissue, particularly in the rectal wall (Teh et al. , 2005; Song et al. , 2007).
The perturbation of dose due to air cavities has been shown to be amplied for smaller
elds (Li et al. , 2000; Martens et al. , 2002). Any dosimetric eects of the rectal
balloon cavity may thus be increased when using intensity modulated radiotherapy
(IMRT) and helical tomotherapy, where multiple small segments are used in place
of the larger open elds seen in 3D conformal and box techniques. Multiple small
segments incident from multiple angles surrounding an air cavity creates an interesting
dosimetric situation. How commercial treatment planning systems calculate the dose
in these situations requires investigation.
3
6.1. Introduction
116
6.1.3 Dose calculation in heterogeneous regions
This paper investigates the convolution/superposition dose calculation algorithm (Mackie
et al. , 1985a) with the collapsed cone convolution method (Ahnesjo, 1989) used by
both the Pinnacle RTPS and the TomoTherapy Hi-Art RTPS . The convolution/superposition algorithm involves the superposition of the energy imparted by primary
photons (usually called the TERMA - Total Energy Released per unit MAss) with
polyenergetic primary and scatter dose deposition kernels. These kernels represent
the dose deposited from primary and secondary radiation around a primary interaction site and are generated using Monte Carlo simulations (Mackie et al. , 1988). For
heterogeneous regions, the kernels are scaled according to the electron density based
on the average density between the primary interaction site and the voxel of interest
(Hoban et al. , 1990). This rectilinear density scaling has been shown to introduce
small errors in the dose calculation in regions of low density, specically lung tissue
(Hoban et al. , 1990; Woo & Cunningham, 1990). These errors may or may not be
observable in the rectal balloon cavity situation and may lead to inaccurate calculation of the dose to the wall of the balloon cavity. The extent of this phenomenon is
examined in this chapter.
6.1.4 Hypofractionation
Recent studies have reported that the / -ratio for prostate cancer is lower than the
conventional value of 10 Gy for tumour (Brenner & Hall, 1999; Fowler et al. , 2001;
Brenner et al. , 2002). Some studies have reported values as low as 1.5 Gy which is
lower than the / -ratio for late rectal complications (/ 3 Gy) (Brenner & Hall,
1999; Fowler et al. , 2001; King & Fowler, 2001; Brenner et al. , 2002; Chappell et al. ,
2004). To maximise the benet of a lower prostate / -ratio than that for late rectal
complications hypofractionation regimes have been introduced (Kupelian et al. , 2001;
117
6.2. Method
Logue et al. , 2001; Kupelian et al. , 2007). In conventional fractionation the prolonged
delivery will lead to repair of the rectal mucosa however with hypofractionation the
same may not apply. In fact some hypofractionated regimes have reported comparable
but slightly higher rectal toxicity rates than standard fractionation (Arcangeli et al.
, 2008; Leborgne & Fowler, 2008). As hypofractionated regimes deliver larger doses
per fraction, possible limitations in the convolution/superposition dose calculation
algorithm must be accurately calculated.
The accuracy to which treatment planning systems calculate dose in the presence of
a rectal balloon cavity must be known to ensure that rectal toxicity data is correlated
with the correct dose. This becomes more pertinent for hypofractionated delivery
regimes. In this phantom study the dose to the rectal wall in the presence of a rectal
balloon was measured using radiochromic lm. The dose distributions were measured
for a 3DCRT plan, an IMRT plan and a helical tomotherapy plan. The results were
compared with calculations from two commercial radiotherapy treatment planning
systems (RTPS).
6.2 Method
6.2.1 Phantom setup
An 8x8x16cm phantom was constructed from acrylic to match the external contours
of an EZ-EM balloon catheter. This is shown in Figure 6.1a. This phantom was
then sandwiched between slabs of solid water and placed between the two halves of
a circular phantom having a 36cm diameter yielding an ellipsoid with a short axis of
36cm and a long axis of 44cm approximating the pelvis anatomy. The rectal balloon
insert was placed in either the lower (sagittal geometry) or upper (spiral geometry)
half of the resulting pelvic phantom. This was then placed in an alpha cradle. The
3
118
6.2. Method
whole setup for the spiral geometry case is seen in Figure 6.1b.
6.2.1.1 Sagittal geometry
The lm was set up in two dierent geometries. The rst, termed 'sagittal geometry',
had the balloon phantom with the two halves aligned such that a sheet of radiochromic
(Gafchromic EBT, International Specialty Products, Wayne, NJ, USA) lm was placed
in the sagittal plane. No balloon was in place for these measurements however the
air cavity created by the Perspex phantom remained, representing the balloon cavity.
The sheet of EBT lm was 15x16cm and covered the region extending over the whole
balloon phantom cross section and 8 cm above the balloon phantom through the target.
The sagittal geometry setup is shown in Figure 6.1c.
Radiochromic lm has been shown to be an excellent dosimeter in cavity situations
(Martens et al. , 2002; Paelinck et al. , 2003, 2005). Paelinck et al. (2003) however
showed that radiochromic lm (Gafchromic MD-55) in a cavity in the central axis of a
beam irradiated edge-on can under-respond at the distal cavity edge due to attenuation
in the lm through the cavity. The under-response was 6-7% for this particular type
of lm and is present only when the lm is in the beam's central axis and irradiated
edge on. This eect was investigated for the Gafchromic EBT lm by comparing the
sagittal geometry setup for a single anterior-posterior beam with and without lm in
the cavity. An under-response of 5.3% was found for the anterior-posterior beam when
there was lm in the cavity.
2
6.2.1.2 Spiral geometry
The second geometry, termed 'spiral geometry' had the phantom in the prone position
with the rectal balloon in place. Three strips of EBT lm were cut and taped together
to give a strip sized 76.2x1.5cm . This was then wrapped around the inated balloon
in a spiral fashion. The length of the lm strip wrapped around the balloon, inated
2
6.2. Method
119
Figure 6.1: Phantom setup a) acrylic phantom to hold EZ-EM rectal balloon catheter
b) full phantom setup in prone position c) schematic diagram showing the location of
the sagittal lm (in blue) d) schematic diagram showing the location of the lm spiral
(black lines wrapping around inside of balloon cavity)
6.2. Method
120
Table 6.1: IMRT and Helical Tomotherapy optimisation parameters
Structure Weight Max Max dose DVH DVH Min Min dose DVH
Dose penalty vol (%) dose dose penalty penalty
PTV
300 70
100
98
98 70
70
2000
Rectum 40
70
150
20
25 700
Bladder 10
70
70
20
40 25
Femurs
5
40
2
20
20 5
with 60cm of air, approximately 4.7 times. This gave a layer of lm 1.2mm thick
around 70% of the balloon diameter. The spiral geometry setup is shown in Figure
6.1d.
3
6.2.2 Treatment plans
A planning CT was taken of the phantom in both geometry setups. In the sagittal
geometry no balloon was in place during image acquisition. For the spiral geometry
the balloon was in place and a 'dummy' lm spiral was put in place for the CT scan
which allowed the lm spiral to be visible on the scan and contoured as a region of
interest (ROI) representing the rectal wall. Seven eld IMRT and 3DCRT Treatment
plans were generated on the Pinnacle RTPS (Philips Radiation Oncology Systems,
Fitchburg, WI, USA) for a Varian Trilogy linear accelerator (Varian, Palo Alto, CA).
Beam angles (IEC convention (IEC, 1996)) of 120, 80, 40, 0, 320, 280 and 240
were used. Helical tomotherapy plans were also generated using the TomoTherapy
Hi-Art planning system, version 2.2.4.1.1, (TomoTherapy Inc, Madison, WI, USA). A
eld width of 2.5cm and pitch of 0.215 was used. Optimisation parameters for IMRT
and helical tomotherapy plans are given in Table 6.1. The planned dose distributions
for the IMRT and helical tomotherapy plans are shown in Figure 6.2.
For each 3DCRT, IMRT and helical tomotherapy delivery three measurements
were performed for both the spiral and vertical geometries. The results presented
121
6.2. Method
are the mean of the three measurements and error bars represent the 95% condence
interval (CI) of the mean. In our case the 95% CI of the mean is obtained by adding
and subtracting the product of the standard error of the mean times the t* value
corresponding to a p-value of 0.05 and two degrees of freedom from the mean. A full
description of this method is presented in Appendix C.1.
The under-response of the lm when irradiated edge-on was investigated for the
measured lms from the treatment plan deliveries. As stated, this eect is only present
for the anterior-posterior beam which carries a beam weighting of 10% for the 3DCRT
and IMRT plans. Therefore an under-response of 5.3% for this eld will lead to an
under-response of 0.53% in the 3DCRT and IMRT deliveries which has been applied
to the posterior rectal wall doses. This will be even less for the helical tomotherapy delivery due to the much greater number of beam angles used. As an exact
anterior-posterior beam weighting is not known for the helical tomotherapy delivery
no correction was applied however this is expected to be a small fraction of the 95%
CI range.
6.2.3 Single elds
So as to understand the eect of the cavity on individual elds, single eld measurements were performed. An anterior-posterior and a lateral eld were investigated. The
lm was in the sagittal geometry with the jaw (8x9cm ) and SSD settings used in the
A-P and lateral elds in the 3DCRT plan. Each eld was irradiated individually with
three lms taken for each separate beam. The results presented are the average of
the three measurements and the error bar represent the 95% condence interval of the
mean.
2
6.3. Results
122
6.2.4 Film calibration
Two sets of calibration lms were taken for the EBT lm, one for the 3DCRT, IMRT
and single eld irradiations and a second set for the helical tomotherapy delivery. The
rst calibration was performed by plotting 5th order polynomial curve to ten dose
points from 0-3.5 Gy that were measured on separate lms on a Varian Trilogy linac.
For the second calibration set a 5th order polynomial curve was tted to nine dose
points from 0-4.3 Gy that were measured on separate lms using the TomoTherapy
beam using the following method. A procedure was set up that delivered radiation
for a given length of time. The dose was measured using an ionization chamber to
obtain the dose delivered for a given beam-on time. Integer multiples of this beam-on
time were then used to deliver varying doses. The EBT lms were scanned on an
Epson Perfection V700 atbed scanner. All lms were scanned at least 24 hours post
irradiation to minimize post-irradiation colour eects (Cheung et al. , 2005)). All
lms were scanned in landscape orientation with a scanning resolution of 75dpi. The
resultant images were 32-bit colour from which the red channel chosen was for analysis.
Background corrections were made to remove spatial non-uniformities by scanning
each EBT lm prior to irradiation. All analysis was performed on a desktop PC using
ImageJ and Matlab software with the Computational Environment for Radiotherapy
Research (CERR, University of Washington in St. Louis) package (Deasy et al. , 2003).
6.3 Results
6.3.1 Sagittal geometry
For both the single eld and treatment plan sagittal lm measurements the measured
and calculated doses at the posterior and anterior cavity walls were compared. The
calculated dose was taken as the dose in the voxel with a CT number between that of
6.3. Results
123
Figure 6.2: Planned dose distributions of the IMRT (left) and helical tomotherapy
plans. The dierences in delivery techniques are seen clearly; IMRT is delivered using
seven beams whereas helical tomotherapy is delivered using multiple smaller beamlets
from the full 360deg
air and water at each border. For the 3DCRT and IMRT plan this was a voxel 2mm
wide in the anterior-posterior direction and for the TomoTherapy plan 1.875mm wide.
The measured dose was taken as the average of all the pixels in the region covered by
the planning system voxel.
The results of single eld irradiations are shown in Figure 6.3. Figure 6.3a shows the resultant digitised sagittal lm image from a single
laterally incident beam. Figure 6.3b shows the resultant digitised sagittal lm image
from a single anterior-posterior beam. Table 6.2 summarises the measured and RTPS
calculated doses to the anterior and posterior rectal wall with and without the cavity.
For the LAT beam the cavity reduced the anterior rectal wall dose and increased the
posterior rectal wall dose. For the LAT beam the RTPS accurately calculated the
anterior rectal wall dose (within 95% CI of the mean) both with and without the cavity. The posterior rectal wall dose was over-predicted with no cavity. This region is
outside of the treatment beam and is dependent on the accuracy of the RTPS model
outside of the eld. In practice doses in this region are of limited importance due to
6.3.1.0.1 Single elds
6.3. Results
124
Table 6.2: Single eld measurements and RTPS calculations of anterior and posterior rectal wall doses with and without rectal balloon cavity. All errors are the 95%
condence interval of the mean.
Beam
anterior cavity wall
posterior cavity wall
EBT Film Plan Di. EBT Film Plan Di.
(Gy)
(Gy) (Gy)
(Gy)
(Gy) (Gy)
LAT - No Cavity 0.530 0.010 0.537 -0.007 0.024 0.011 0.038 -0.014
LAT - Cavity 0.468 0.053 0.514 -0.046 0.053 0.003 0.046 0.007
AP - No Cavity 0.790 0.060 0.786 0.004 0.570 0.020 0.582 -0.012
AP - Cavity 0.776 0.049 0.784 -0.008 0.625 0.011 0.643 -0.018
AP - Cavity (no 0.785 0.006 0.784 0.001 0.659 0.004 0.643 0.016
lm in cavity)
their low value. The posterior rectal wall dose was under-predicted when the cavity
was present. For the single AP beam the cavity had no eect on the anterior rectal
wall dose but increased the posterior rectal wall dose. For the AP beam the RTPS
calculated the anterior rectal wall dose within the 95% CI of the measurement both
with and without the cavity. The posterior rectal wall dose was calculated within the
95% CI of the measurement with no cavity but was over-predicted when the cavity was
present. When the lm was removed from the cavity and placed only on the outside
for a single AP beam, the eect of the attenuation in the lm through the cavity is
evident; the posterior rectal wall dose increased relative to the measurement with the
lm in the cavity. When compared to the RTPS calculation, the measured dose at the
posterior rectal wall was higher, showing an under-prediction by the RTPS.
Figure 6.4 shows the digitised lm images and the
resultant dose proles taken in the anterior-posterior direction for the 3DCRT, IMRT
and helical tomotherapy plans. In all three treatment techniques the rectal balloon
cavity was seen to perturb the dose distribution. For all three plans the anterior cavity
wall dose was over-predicted and the posterior cavity wall dose under-predicted by the
relevant planning systems. The measured and calculated doses are given in Table 6.3.
6.3.1.0.2 Treatment plans
6.3. Results
125
Figure 6.3: Sagittal lm results from (a) single laterally incident beam and (b) single
anterior-posterior beam with and without a cavity. The white lines show the location
of the proles. The arrows show the beam direction. Horizontal error bars on the plan
data show the width of the planned dose voxels.
Table 6.3: Measured and planned cavity wall doses. Percentage dierences are
measured-planned normalized to measured dose. Errors quoted are the 95% condence interval.
Technique Anterior Cavity Wall Posterior Cavity Wall
EBT Film Plan Di. EBT Film Plan Di.
(Gy) (Gy) (Gy) (Gy) (Gy) (Gy)
3DCRT 68.4 1.1 69.8 -1.4 19.2 1.0 16.4 2.8
IMRT
65.8 2.5 69.7 -3.9 13.3 0.9 11.2 2.1
Helical 67.3 1.7 70.0 -2.7 11.6 1.1 6.9 4.8
Tomotherapy
126
6.3. Results
6.3.1.1 Spiral geometry
The spiral lm strips, which represent a surrogate
for the rectal wall, were scanned and converted into absolute dose. Two analyses were
performed. In the rst, a line prole was taken across the centre 1cm of the lm along
the length of the spiral. This averaged all of the pixels across the central 1cm of the
lm strip at each point along the length of the lm. The dose projection tool in CERR
was then used to get the average dose across the lm contour in the superior-inferior
direction, at each point around the rectal balloon cavity. The measured and calculated
doses were then plotted as a function of angle around the cavity. The outermost loop
and innermost full loop only were plotted for clarity. These constitute the two extremes
of the dose gradient across the thickness of the lm spiral. The outermost loop is the
dose to a 0.040mm thick strip centred at 0.117mm from the outside of the balloon
cavity and the innermost loop is the dose to a 0.040mm thick strip centred 0.819mm
from the outside of the cavity. The planning system represents the dose averaged over
a 1.5mm thick voxel around the outside of the cavity. Each loop of the lm spiral is
thus measuring a dierent point contained within the lm ROI. The measured and
calculated proles are shown in Figure 6.5a.
Dose-volume histograms (DVHs) were generated from the digitised lm images.
This was done by obtaining histograms of the digitised lm images and normalizing
the volume approximating the pixels as volume elements. This is valid since an individual pixel can be considered the dose measured in a voxel whose dimensions are the
resolution of the digitised lm image (0.034x0.034cm at 75dpi) by 40m (the thickness of the active layer in Gafchromic EBT lm). The measured voxels constitute a
fraction of the total volume of the lm spiral however due to the geometry the voxels
can be considered as a representative volume of the total lm spiral. The lm spiral
was visible on the CT scans as a layer one CT voxel thick (1.5mm) around the inside
6.3.1.1.1 Treatment Plans
2
6.3. Results
127
Figure 6.4: Sagittal digitised lm images and resultant dose proles for a) 3DCRT
b) IMRT and c) helical tomotherapy (HT) delivery techniques. The colour bar is in
absolute dose in Grays. All measurements were scaled to represent the dose delivered
over the total treatment (28 fractions). The error bars are the standard error of three
measurements.
128
6.3. Results
of the cavity. This was contoured as the lm spiral ROI. The volume of the lm spiral
ROI (3.42cm ) was dierent to that of the actual lm spiral (2.68cm ) due to the size
of the CT scan voxels. The dierences in the volume were due to dierences in dimensions between the actual lm spiral and the lm spiral ROI, primarily in the thickness
of the lm spiral (the actual lm spiral is 1.2mm thick and the lm spiral ROI is
1.5mm thick). The actual lm spiral was contained completely by the lm ROI. There
is some uncertainty in the location of the lm spiral within the lm ROI. This may
lead to discrepancy between the measured and planned dose that depends on whether
the lm spiral is at the centre, outside or inside of the lm ROI. This dierence in
measurement location is < 0.3mm and as such will be within the experimental error of
the three measurements. The measured DVHs were compared with those calculated
by the planning systems for the lm spiral region of interest (ROI). The volumes were
normalised to percentage volumes to aid comparison. The measured versus planned
DVHs are shown in Figure 6.5b.
Both the 3DCRT and IMRT planned DVHs match their respective measured DVHs
in the lower dose region (< 35Gy). However for doses >35Gy the measured volumes
for doses up to 70Gy are less than planned. Slight dierences exist in the dose range
15-20Gy which represents the dose at the posterior cavity edge. The measured doses
were higher than the planned doses as seen in Figure 6.4 which is then represented
by higher volumes receiving 15-20Gy. For both the 3DCRT and IMRT plan large
discrepancies were found between the planned and measured DVH for doses between
60 and 70Gy. The measured DVH shows much lower rectal wall volumes receiving
dose between 60 and 70Gy.
For the helical Tomotherapy plan discrepancies occurred between the measured and
planned DVH in the dose region between 15Gy and 72Gy, where the measured volumes receiving a given dose were more than the planned volumes by varying amounts.
3
3
6.4. Discussion
129
Table 6.4: Measured and planned rectal wall percentage volumes receiving specied
doses. Reported error is the 95% condence interval of the mean of three measurements.
Parameter
3DCRT
IMRT
HT
EBT Film Plan EBT Film Plan EBT Film Plan
V25 (%) 67.7 0.4 67.6 55.7 2.5 54.1 71.9 0.5 58.8
V50 (%) 39.1 0.4 42.3 32.7 1.7 35.9 41.1 0.9 38.3
V60 (%) 32.9 0.6 36.9 18.6 1.9 26.7 26.2 1.7 22.9
V65 (%) 12.6 0.8 32.7 4.3 2.9 21.2 21.2 1.5 17.0
V70 (%) 0.00 0.01 10.5 0.0 0.0 6.7 14.6 0.6 11.6
Volumes receiving doses > 72Gy were accurately calculated. The V25, V50, V60, V65
and V70 values for the three treatment techniques are summarised in Table 6.4. It
should be noted that the doses to the rectum were higher in the helical tomotherapy
plan than in the 3DCRT or IMRT plan, due to a too-low weighting on the maximum
PTV dose objective. This does not impact on the results however, as the goal is to
determine the accuracy of the dose calculation around the rectal balloon cavity.
6.4 Discussion
In this report the dosimetric eect of a rectal balloon cavity and the accuracy two
commercial RTPSs in calculating the dose around the cavity was investigated. Dose
perturbation for single lateral and anterior-posterior elds were initially investigated.
This was then extended to 3DCRT, IMRT and helical tomotherapy delivery techniques.
6.4.1 Single elds
For the lateral beam the impact of the cavity is to reduce
the dose to the anterior rectal wall and increases the dose to the posterior rectal wall.
The posterior rectal wall dose was increased due to the greater electron range through
the cavity leading to a higher uence of lateral electrons at the posterior rectal wall.
6.4.1.0.2 Lateral eld
6.4. Discussion
130
The RTPS accurately models the dose to the anterior rectal wall within error but in
the measurements there is a clear trend of decreasing dose in the 2mm proximal to the
anterior cavity wall that is not seen in the RTPS calculation. At the posterior rectal
wall the RTPS under-predicted the dose.
There are similarities between these lateral eld results and other reports investigating head and neck and lung cavities. The phenomenon of lateral electron disequilibrium has been experimentally characterized in lung by several investigators. The
main observations for small elds in lung are dose voids in the central axis and small
secondary build up regions beyond the lung tissue interface (Mackie et al. , 1985b;
Metcalfe & Battista, 1988; Metcalfe et al. , 2007). For larger elds while central axis
dose voids are reduced, penumbral aring is still an observable phenomenon (Kornelsen & Young, 1982; Young & Kornelsen, 1983). With the relatively large elds
used in this study there was only a slight dose reduction at the anterior cavity wall
but penumbral aring was inferred by the decreased anterior rectal wall dose and increased posterior rectal wall dose. The penumbral aring is because the electron range
extends in low density regions, causing the penumbral width to broaden. The eect of
lateral electron disequilibrium on dose beyond air cavities in the head and neck region
has been discussed (Massey, 1962; Nilsson & Schnell, 1976; Wong et al. , 1992). In
this report we have not investigated higher energy photon beams but previous studies
in lung cavities suggest that penumbral aring will increase with energy (Kornelsen
& Young, 1982; Young & Kornelsen, 1983; Metcalfe et al. , 1993). As higher energies
such as 10MV and 18MV are often used for lateral elds the penumbral aring eect
may increase in these cases.
Another aspect that warrants discussion is the CT numbers within the balloon
cavity. Some CT reconstruction algorithms have limitations that result in air cavities
within a patient/phantom having dierent (slightly higher) CT numbers than that for
6.4. Discussion
131
Figure 6.5: Measured and planned rectal wall doses and resultant DVH from spiral
lm geometry. (a) represents the dose to the outermost and innermost loop of the lm
spiral and the planned dose to the lm spiral for the 3DCRT plan (d) represents the
resultant rectal wall DVH from the lm spiral and the planned rectal wall DVH for
the 3DCRT plan. (b) and (e), and (c) and (f) represent the same for the IMRT and
helical tomotherapy plans respectively.
6.4. Discussion
132
air outside of the patient. An over-estimate of the CT number in the cavity would
lead to an under-estimate of the disequilibrium conditions for the dose calculation. In
this specic case, the CT numbers for the air in the cavity and the air outside of the
phantom were approximately equal, however for other air cavities, CT scanners and
reconstruction algorithms, this may not be the case. In these cases, it is advisable to
monitor the CT numbers of air cavities within patients to possibly predict where dose
calculation limitations may exist.
For the anterior-posterior beam the cavity
had no eect on the anterior rectal wall but increased the posterior rectal wall dose.
No build-down eect was seen, as observed by Li et al. (2000). This may be due to the
relatively large eld size used here compared with the cavity size. The increased posterior rectal wall dose was due to the reduced attenuation through the cavity meaning
higher photon uence at the posterior edge of the cavity. A slight secondary build-up
is seen distal to the posterior cavity edge, however this eect is minimal due again to
the relatively large eld size used and the depth of the cavity. This is in agreement
with Li et al. (2000) who found a secondary build-up at the distal cavity edge whose
magnitude was reduced with increased eld size and cavity depth. The RTPS calculated the dose to the anterior rectal wall within the 95% CI of the measurement but
over-predicted the dose to the posterior rectal wall.
6.4.1.0.3 Anterior-posterior eld
6.4.2 3DCRT, IMRT and helical tomotherapy deliveries
Multiple elds and segments were combined in the 3DCRT, IMRT and helical tomotherapy plans. With the sagittal lm geometry both the Pinnacle RTPS and TomoTherapy RTPS over-predicted the anterior rectal wall dose (by 1.43Gy, 3.92Gy and
2.67Gy for 3DCRT, IMRT and helical tomotherapy respectively) and under-predicted
the posterior rectal wall dose (by 2.62Gy, 2.01Gy and 4.79Gy for 3DCRT, IMRT and
6.4. Discussion
133
helical tomotherapy respectively). These two eects are similar to that seen in the
single lateral eld irradiation which is expected, since the majority of the radiation for
the three plans was delivered from angles oblique to the anterior cavity wall.
For the spiral lm geometry for the 3DCRT and IMRT plans, the Pinnacle RTPS
calculated dose was seen to agree with the measured dose to the outermost loop in the
lm spiral with the exception of the dose to the anterior 60 of the cavity wall. The
Pinnacle RTPS over-predicted the dose to the anterior 60 of the cavity wall. When
compared with the innermost loop of the lm spiral the Pinnacle RTPS over-predicted
the dose to the anterior 60 of the cavity wall and under-predicted the dose to the
posterior 120 of the cavity wall. These results were reected when the lm spiral was
converted to a DVH and compared with the lm ROI DVH calculated by the planning
system. The Pinnacle RTPS accurately calculated the volumes receiving < 35Gy but
over-predicted the volumes receiving > 35Gy.
The accuracy of the TomoTherapy RTPS when compared with lms in the spiral
geometry varied around the cavity. It should be emphasised that although higher doses
are being delivered to the rectal wall than in the 3DCRT and IMRT plans, we are not
judging the quality of the plan, but the agreement between the RTPS calculated and
the measured dose. The RTPS calculated anterior cavity wall dose lies between that of
the outermost and innermost loops of the lm spiral; once the planned doses and lm
spiral doses are converted to a DVH this averages out giving an accurate calculation
of the volumes receiving high doses. The TomoTherapy RTPS then under-predicts the
intermediate and low doses, leading to the reduced volumes receiving these doses in
the DVH.
134
6.4. Discussion
6.4.3 Clinical signicance
It is clinically signicant that for the 3DCRT and IMRT plans the V70, V65, V60
and V50 values are over-predicted by the Pinnacle RTPS. The V70, V65, V60 and
V50 parameters are correlated with incidence of rectal bleeding (Jackson et al. , 2001;
Fiorino et al. , 2002; Huang et al. , 2002b). Any reduction in these parameters should
result in a decreased incidence of rectal bleeding under that predicted by the planning
system. This has been observed in other studies (Sanghani et al. , 2004; D'Amico et al.
, 2006; van Lin et al. , 2007). Conversely, the TomoTherapy RTPS under-predicted
the V70, V65, V60 and V50 values but accurately predicted volumes receiving higher
doses than 70Gy. Any under-prediction of these values could lead to an unexpected
increase in rectal toxicity, particularly if delivering a hypofractionated schedule. Given
this observation one may want to consider a reduction in dose volume objectives placed
on the rectal wall when employing helical tomotherapy.
The results from the spiral lm geometry suggest that the Pinnacle RTPS overestimates the dose to the anterior rectal wall, but the TomoTherapy RTPS is accurately
calculating or slightly under-predicting the anterior rectal wall dose. The consequence
of this is that dose volume constraints for the rectal wall acquired from 3DCRT and
IMRT studies may have been over-estimated, where as for tomotherapy they might
have been correct or even under-predicted. Any dose-volume constraints deemed 'safe'
from 3DCRT and IMRT studies that have subsequently been applied to tomotherapy
cases may need to be reconsidered.
The eect of the rectal balloon cavity on treatment plan delivery has been investigated in other reports (Teh et al. , 2005; Song et al. , 2007) but the accuracy of
the convolution/superposition algorithm has not been investigated with regard to the
rectal balloon cavity. Teh et al. (2005) measured the dose to the rectal cavity wall
for a single 2x2cm eld as well as a serial tomotherapy delivery. A 15% reduction in
2
6.5. Conclusions
135
the dose to the anterior cavity wall due to the presence of the rectal balloon cavity
was observed. Song et al. (2007) used Monte Carlo simulations to investigate the
accuracy of the Eclipse RTPS (with no heterogeneity corrections) in the presence of a
rectal balloon. The Eclipse RTPS was found to over-predict the volumes receiving >
96% of the prescription dose and an under-prediction of the rectal volumes receiving
< 22% of the prescription dose. It was determined that the dierences at these two
ends of the dose range were due to inaccurate calculation of the dose from two lateral
beams. The results presented in this investigation agree with these two reports.
The important issue is that if heterogeneity correction is used in the RTPS then
the convolution/superposition algorithms (Boyer & Mok, 1985; Mackie et al. , 1985b;
Mohan et al. , 1986) used by both Pinnacle and TomoTherapy RTPSs eectively model
these disequilibrium situations. These models suer from some small electron range
scaling issues (Keall & Hoban, 1996) that can lead to inaccurate modelling of cavity
interface doses as seen in this report. The magnitude of these inaccuracies in the high
dose region is small but may be clinically signicant considering evidence suggests most
disease is likely to be found in the transitional zone, that is, in the lobes and close to
the rectal border (Chen et al. , 2000). They do however accurately show the qualitative
eect of this disequilibrium region. The utility of Monte Carlo simulations in these
situations becomes apparent (Song et al. , 2007). Monte Carlo accurately models
disequilibrium situations and may provide clinicians with more precise dosimetry. The
importance of in vivo dosimetry also becomes evident; accurate measurements of the
dose received by the rectal wall will provide more accurate data for toxicity correlations.
6.5 Conclusions
This report details the eect of an air cavity created by a rectal balloon and the
accuracy of two commercial treatment planning systems in calculations surrounding
6.5. Conclusions
136
the cavity. When irradiated with single elds of the same size as that seen clinically,
the cavity was seen to perturb the dose at the cavity walls. For a single lateral beam
the cavity lead to a decrease in the anterior rectal wall dose and an increase in the
posterior cavity wall dose. This was due to penumbral aring through the air cavity.
For a single anterior-posterior beam the cavity was seen to increase the posterior dose.
The Pinnacle RTPS predicted the qualitative eects of the cavity but under-estimated
the eect of the cavity. For clinically relevant treatment plan delivery, the Pinnacle
and TomoTherapy RTPSs both over-predicted the anterior rectal wall dose and underpredicted the posterior cavity wall dose for 3DCRT, IMRT and helical tomotherapy
deliveries. This was visible on the sagittal lm geometry. For the spiral lm geometry
the Pinnacle RTPS was seen to over-predict the high dose region at the anterior rectal
wall. The dose to the posterior rectal wall was under-predicted by the Pinnacle RTPS.
The TomoTherapy Hi-Art RTPS under-predicted the low and intermediate doses to
the rectal wall but accurately calculated the high dose region at the anterior cavity wall
adjacent to the prostate. If the Pinnacle RTPS over-predicts but the TomoTherapy
RTPS accurately calculates the anterior rectal wall dose then dose volume constraints
carried into tomotherapy treatments from 3DCRT and IMRT treatments may need to
be reconsidered
Chapter 7
On the feasibility of in
vivo
real-time rectal wall dosimetry for
prostate radiotherapy
7.1 Introduction
Modern prostate cancer external beam radiotherapy involves the delivery of high doses
using highly conformal delivery techniques. Increased prostate dose has been shown
to increase local control (Hanks et al. , 1998; Zelefsky et al. , 1998; Pollack et al. ,
2000). Additionally, hypofractionation has been shown to be an attractive delivery
method due to the apparent low alpha/beta ratio for the prostate (Brenner & Hall,
1999; Fowler et al. , 2001; Brenner et al. , 2002). Increasing the total treatment dose
or the dose per fraction increases the need for accurate delivery verication.
In Chapter 6, it was seen that there was some uncertainty in the rectal wall dose in
the presence of a rectal balloon. It was suggested that in vivo dosimetry of the rectal
wall could be an attractive method for verication of target and organ at risk (OAR)
doses. An in vivo dosimeter placed on the anterior rectal wall allows the clinician to
137
7.2. Methods and materials
138
measure the dose delivered to the section of the rectum receiving the highest dose. This
would be useful in the context that the planning system calculation may be innacurate
at the cavity wall. In addition to this, the anterior rectal wall is generally contained
by the Planning Target Volume (PTV) therefore any dosimeter placed on the anterior
rectal wall can be used as a surrogate for the PTV dose at the posterior region of the
volume.
Rectal balloons are used in a number of institutions in prostate radiotherapy to
immobilise the prostate and reduce the amount of rectal wall irradiated to high doses
(McGary et al. , 2002; Teh et al. , 2002; Wachter et al. , 2002; Patel et al. , 2003; Teh
et al. , 2005; van Lin et al. , 2005b,a, 2007). Rectal balloons are placed in the rectum
for each treatment fraction. Rectal balloons provide an excellent means for placement
of an in vivo dosimeter.
MOSFET dosimeters have been used extensively for in vivo dosimetry (Butson
et al. , 1996; Scalchi & Francescon, 1998; Quach et al. , 2000; Scalchi et al. , 2005;
Zilio et al. , 2006). The advantages of MOSFET detectors for point dose measurements
are their relatively small size and real-time readout capabilities.
This phantom study investigates the feasibility of using a commercial rectal balloon in combination with a novel MOSFET detector to obtain real-time in vivo dose
measurements of the rectal wall. Real-time rectal wall measurements were carried out
on a specially designed phantom in a helical tomotherapy treatment scenario.
7.2 Methods and materials
7.2.1 MOSFET measurements
A novel MOSFET detector, the MOSkin, developed at CMRP, was used for all measurements. The MOSkin has been described in detail elsewhere and has been used
7.2. Methods and materials
139
Figure 7.1: The MOSkin detector placed on the RadiaDyne rectal balloon
extensively for in vivo measurements (Kwan et al. , 2007, 2008b,a, 2009; Qi et al. ,
2009). In summary, the MOSkin has reproducible Water Equivalent Depth (WED)
of measurement of 70m. This makes the MOSkin ideal for dosimetry in high dose
gradient regions, such as the rectal wall in the presence of a rectal balloon air cavity.
Two MOSkins were used for the measurements. The MOSkins were calibrated in a
6MV, 10x10cm eld at 100cm SSD at 10cm depth in water using a Varian 600c linear
accelerator (Varian Medical Systems, Palo Alto, CA, USA).
A commercial rectal balloon was used for all measurements (RadiaDyne, Houston
TX, USA). A custom made phantom was built to match the external contours of the
balloon. This phantom was placed in a pelvic phantom to simulate a human pelvis. A
planning CT scan was taken of the phantom set up. The CT data was transferred into
the Pinnacle RTPS (Philips Medical Systems, Middleton, WI, USA) where contours of
a hypothetical prostate, rectum (the balloon cavity), bladder and femoral heads were
created. The CT and contour data was then transferred to the TomoTherapy RTPS
(TomoTherapy, Madison, WI, USA) on which a helical tomotherapy plan was created.
2
7.2. Methods and materials
140
Table 7.1: Helical tomotherapy optimisation parameters. All doses are in Gy.
Max Max dose DVH DVH Min Min dose DVH
ROI Weight Dose
Penalty Vol (%) Dose Dose Penalty Penalty
PTV 300 70
100
98
70 70
70
2000
Rectum 40
70
150
20
25
700
Bladder 10
70
70
20
40
25
Femurs 5
40
2
20
20
5
A eld with of 2.5cm, pitch of 0.215 and optimisation parameters given in Table 7.1
were used. The plan and dose distribution was imported into the Computational Environment for Radiotherapy Research (CERR, University of Washington in St. Louis,
MO, USA) (Deasy et al. , 2003) platform for analysis.
Two measurements were performed. In the rst, a single MOSkin was attached
to the outside of the balloon at the anterior most location, shown in Figure 7.1.
This corresponds to the location between the balloon and the anterior rectal wall.
The balloon was then placed in the phantom setup and a single 2.5Gy fraction was
delivered. This was repeated twice.
In the second measurement, six locations around the balloon, spaced every 60
around the balloon starting at the anterior most point, were selected as measurement
points. The locations of the detectors are given in Figure 7.2. The MOSkin detectors
were placed around the outside of the balloon cavity, facing outwards such that the
distance between the cavity wall and the MOSkin sensitive layer was 70m WED.
The MOSkins were irradiated in separate deliveries. No balloon was in place for this
measurement.
For both MOSkin measurements the MOSkin was powered by a CMRP-developed
reader which was then connected to a laptop computer. The readout of the MOSkin
was carried out using in-house developed software, MOSPLOT. MOSPLOT acts as a
controller for the MOSFET reader. This software allows reading of the MOSkin at
7.3. Results
141
designated time intervals. The MOSkin was read out at 1Hz for the entire duration of
the radiation delivery for all measurements.
7.2.2 Radiochromic lm measurements
In addition to the MOSkin measurements, a single piece of Gafchromic EBT lm
was placed at the anterior rectal wall between the balloon and the cavity wall and a
single fraction was delivered. This was repeated twice. This was done to provide a
comparison with the rst set of MOSkin measurements. The lm was calibrated in a
tomotherapy beam according to the procedure outlined in a previous chapter (Chapter
6). Briey, separate lms were irradiated from 0-4.3Gy in a tomotherapy beam at 5cm
depth in water. A fth order polynomial was then tted to the calibration points and
applied to the measured lms. All lms were scanned on an Epson Perfection V700
atbed scanner, 24 hours post-irradiation to avoid post-irradiation colouration eects.
Films were scanned in the landscape direction and the red channel was chosen for
analysis.
All measurements were compared with the TomoTherapy Hi-Art RTPS calculation.
The error bars on the MOSkin and EBT lm measurements are the 95% condence
interval of the mean. This was obtained by adding or subtracting the product of the
standard error and the t* value corresponding to a p-value of 0.05 and 2 degrees of
freedom.
7.3 Results
Figure 7.3 shows the results of the single MOSkin measurement located on the anterior
rectal wall. The dose delivered to the MOSkin increases slowly between the start of
the delivery and 130s into the treatment. As the couch motion moved the MOSkin
into the tomotherapy fan beam the dose increased rapidly between 130s and 210s. The
7.3. Results
142
Figure 7.2: Location of MOSkin detectors around the rectal balloon cavity for the
second set of measurements
7.3. Results
143
Table 7.2: Measurement results for anterior rectal wall measurement
Source Measurement (cGy) 95% CI
Plan
261.5
MOSkin
224 3
EBT Film
236 3
Figure 7.3: Anterior rectal wall planned dose compared with measured dose over the
duration of the fraction delivery. Note the dose to the MOSkin is accrued over the
total fraction delivery time.
MOSkin then moved out of the fan beam and the dose entered a plateau region. Table
7.2 summarises the measured and calculated total doses at the end of the delivered
fraction. The MOSkin measurement was 37.95cGy less than the planned dose and
12.62cGy less than the lm measurement.
Figure 7.4 shows the MOSkin measured doses at the six investigated locations (see
Figure 7.2) around the rectal balloon. Figure 7.4a shows the total MOSkin measured
doses compared with the planned doses at the six locations. The total doses are
presented in Table 7.3. The MOSkin measured doses were up to 13% lower than
the planned doses at the anterior locations. As the detector location became more
7.3. Results
144
Table 7.3: Measured and planned doses at the six locations given in Figure 7.2.
Location Planned Dose (cGy) Measured Dose (cGy) Dierence (cGy) % dierence
0
257
232
25
9.7
60
189
178
11
5.8
120
83
80
3
3.6
180
60
62
-2
-3.3
240
98
94
4
4.1
300
200
174
26
13.0
posterior, the dierence between the measured and planned doses decreased.
The MOSkin exhibited some under response when compared to the planned dose
and the dose measured by the EBT lm. We believe this is due to a combination of
two eects. The rst is an over-prediction of the anterior rectal wall dose due to the
balloon cavity, as found in Chapter 6. The second eect is an under-prediction of the
MOSkin measured dose compared with the EBT lm measure dose due to angular
response of the MOSkin detector. The MOSkin detector's asymmetric construction
means that there is an over-response to radiation from the top 180, probably due to a
dose enhancement eect from the a thin Aluminium contact layer near the detector's
front surface. This layer attenuates photons and increases the number of secondary
electrons crossing the gate oxide . There are two correction strategies for the angular
response under investigation - a ltering method and a dual MOSkin method.
7.3.1 Angular dependence correction method 1: Filtering method
The rst correction strategy implemented to account for this was proposed by Rosenfeld (2009b) and investigated theoretically by Lian (2009). The ltering method involves placing a thin layer of Copper on the top side of the MOSkin, in an eort
to decrease the sensitivity of the top side of the MOSkin to match the back side of
the MOSkin. This method was investigated experimentally by using 30m and 60m
7.3. Results
145
Figure 7.4: (a)MOSkin measured rectal wall doses over time for the six investigated
locations around the rectal wall as given in Figure 7.2 and (b) Temporal dose accumulation for the six locations
146
7.3. Results
thick layers of Copper attached to the top side of the MOSkin. The theoretical optimal
thickness was calculated to be 25m for a 10x10cm 6MV photon eld however copper
of this thickness was not readily available at the time of measurement. The Copper
covered MOSkin was placed at depths of 1.5cm and 10cm in solid water and irradiated
with a 5x5cm 6MV photon eld in face up and face down orientations. The same
dose was delivered for both orientations. Figure 7.5 shows the response of the face
up and face down MOSkin at the two depths. With no Cu lter, at 1.5cm depth the
MOSkin response to radiation incident from the back side is 14.7% less than when the
radiation is incident from the top side. With a 30m layer of Cu on the top side, the
responses are within 2%, with the MOSkin being more sensitive to radiation incident
from the back side. With a 60m layer of Cu on the top side, the MOSkin is even more
sensitive to radiation from the back side. It is probable that the single 30m layer was
an over-correction of the responses, which was accentuated by the 60m layer. The
response at 10cm depth is similar to that at 1.5cm depth, however there is a greater
over-correction.
The results observed in Figure 7.5 suggest that the thickness of the lter depends
on the material of the lter and the depth of measurement. The radiation quality
changes depending on the depth of measurement, therefore a lter material dierent
than Silicone is not ideal. Copper was selected as only a small thickness is required.
Ideally, the lter should be made of Silicone and the dimensions should match that
of the Silicone substrate on the back side of the MOSkin. This would give depthindependent correction. The full angular response was not measured for the ltered
detector set up.
2
2
7.3. Results
147
Figure 7.5: Relative response for face up and face down MOSkin orientations with one
and two layers of CU on the top edge at (a) 1.5cm and (b) 10cm depth in solid water.
The error bars represent the 95% condence interval of the mean.
7.3. Results
148
7.3.2 Angular dependence correction method 2: Dual MOSkin
conguration
The second method investigated to correct/remove the angular dependence of the
MOSkin was to combine two MOSkin detectors, placed face-to-face. This method was
rst proposed by Rosenfeld et al. (2005). If detector one (D1) and detector two (D2)
are placed face-to-face at depth, then if D1 is being irradiated from the top, D2 is
being irradiated from the bottom, and vice versa. That is, when the dual detector is
irradiated from any direction, D1 and D2 are over or under responding compared to
the mean. By taking the average of the two detectors, the over and under response is
cancelled out. This is an active correction of the asymmetrical response of the MOSkin
and was proposed by Rosenfeld (2009a).
The dual MOSkin conguration was tested experimentally by placing two MOSkins
together, face-to-face. The dual MOSkin was then placed in a cylindrical phantom that
was embedded in a square, solid water phantom at a depth of 5cm. The dual MOSkin
was then irradiated from a beam incident every 30 for a full rotation. This was
repeated twice. The cylinder containing the dual MOSkin was rotated, rather than
the linac gantry, to maintain constant radiation path length to the location of the
detector. The response to the same dose was recorded for both MOSkin detectors.
The average response of the two detectors was then taken.
Figure 7.6a shows the angular response of the two detectors in the dual MOSkin
setup. The response of the two detectors is 180 out of phase. Figure 7.6b shows the
average response of the two detectors, which is within 2%.
A further set of measurements with the dual MOSkin was performed in using a
clinical IMRT verication phantom - I'mRT (IBA Dosimetry). The dual MOSkin was
placed at the centre of the phantom and calibrated. Calibration was performed by
irradiating the dual MOSkin with a known dose and recording the response of both
7.3. Results
149
Figure 7.6: (a) The response of the two detectors in the dual MOSkin setup. Error
bars (no end cap for D1) are the 95% CI of the mean (b) The average response of the
two detectors. Error bars are the 95% CI of the mean.
7.3. Results
150
detectors. The dual MOSkin was then ipped and the calibration was repeated. For
each calibration measurement the change in threshold voltage for each detector was
added. The average of the summed responses (i.e. in both orientations) was then
divided by the known dose (100cGy) to give the calibration factor. The detector
responses were summed so that any uctuations in read out result in a lower relative
deviation to the total measurement, reducing total error.
The dual MOSkin was then irradiated from gantry angles every 45 for the full
360 (detector was at machine isocentre). The experimental setup is given in Figure
7.7(a). Each measurement was multiplied by the dual MOSkin calibration factor to
result in absolute dose for each measurement. The dual MOSkin was then replaced
with the ion chamber. The ion chamber was calibrated by irradiating with a known
dose. The ion chamber was then irradiated with the same gantry angles as for the dual
MOSkin. The dual MOSkin and ion chamber results were then normalised to their
respective measurements at 0 gantry angle. The dual MOSkin normalised dose was
normalised to the normalised ion chamber measurement to obtain the relative dose for
each incident beam angle. All measurements were repeated twice.
Figure 7.7(b) shows the normalised dual MOSkin response to beams from multiple
angles at the centre of the phantom. With the exception of one angle (225), the dual
MOSkin is within 4% of the ion chamber measurement. The measurement at 225
may have been aected by the beam being incident on the detectors through one of
the metal couch support struts. The dierences in the detection volume of the two
detectors may have led to a discrepancy between the measured doses at this angle.
The dosimetric accuracy of the dual MOSkin coupled to a rectal balloon was then
veried for two treatment plans. A dual MOSkin was attached to the anterior wall of
a rectal balloon. The balloon and MOSkin were placed in the custom made balloon
phantom, which was then inserted into the I'mRT phantom. A planning CT scan was
7.3. Results
151
Figure 7.7: (a) The I'mRT phantom setup for dual MOSkin and (b) The normalised
measurement (dual MOSkin / ion chamber) for each incident beam angle. The error
bars are the 95% interval of the mean for three measurements.
7.4. Discussion
152
taken of the phantom setup. 3DCRT and IMRT plans were created in the Pinnacle
RTPS, delivering 78Gy in 2Gy fractions to a hypothetical prostate target in the phantom. Three fractions each of the 3DCRT and IMRT plans were then delivered to the
phantom, and the dose delivered to the dual MOSkin was recorded at a frequency
of 1Hz during each delivery. The average MOSkin measured doses compared to the
Pinnacle RTPS calculation are presented in Figure 7.8.
Figure 7.8 shows that for the 3DCRT plan the MOSkin measured dose is 2.62%
lower than the planned dose and for the IMRT plan the MOSkin measured dose is
3.17% lower than the planned dose. From Chapter 6 it was shown that the Pinnacle
RTPS over-predicts the anterior rectal wall dose; the results in Figure 7.8 are consistent
with this nding.
7.4 Discussion
This study has presented initial results from testing of a novel MOSFET dosimeter developed at CMRP combined with a rectal balloon for use in real-time in vivo
dosimetry. In this phantom study, the performance of a dual MOSkin conguration
in combination with a commercial rectal balloon was investigated to determine the
feasibility of this apparatus for clinical use.
The MOSkin detectors showed a reproducible change in Vt for doses 0-10Gy. The
MOSkin is thus suitable for in vivo real-time dose monitor for in high dose-per-fraction
hypofractionation schedules. The dual MOSkin has excellent angular dependence of
within 2.5% from all incident beam angles. This makes it particularly suitable for rotational therapies such as tomotherapy and volumetric modulated arc therapy (VMAT).
When placed on the anterior wall of a commercial rectal balloon and irradiated
with clinical 3DCRT and IMRT plans, the dual MOSkin provided real-time dose measurement of the anterior rectal wall dose. The anterior rectal wall dose was measured
7.4. Discussion
153
Figure 7.8: The dual MOSkin measured dose compared with the planned dose for (a)
3DCRT plan and (b) IMRT plan. The error bars represent the 95% condence interval
of the mean of three measurements.
7.4. Discussion
154
to be 2.62% and 3.17% lower than the RTPS calculation for the 3DCRT and IMRT
plans respectively. It was shown in Chapter 6 that the collapsed cone convolution/superposition dose calculation algorithm used in this study over-predicts the anterior
rectal wall dose when an air-lled rectal balloon is used. Therefore the lower measured
dose, relative to the RTPS calculation, concurs with this previous nding.
The MOSkin was visible on the planning CT scan therefore accurate localisation of
the detector at the time of planning is ensured. This allows knowledge of the expected
dose to the detector during delivery. The MOSkin would also be visible on KV conebeam CT or KV portal images, so daily localization of the detector is possible, to
integrate with any adaptive radiotherapy or advanced image guidance protocols.
Although the rectal balloon reduces prostate motion signicantly, rectal motion
can still occur which may increase rectal wall dose. Conversely, if the rectal wall dose
is signicantly lower than the expected dose this could mean part of the PTV is being
underdosed. At the very least, this in vivo dosimetry system provides tracking and
verication of daily delivered rectal wall doses, which may prove useful in the case of
observed enhancement of toxicity or local recurrence. Correlation of rectal toxicity
rates with absorbed dose currently relies on calculated dose at the time of planning.
This device would provide a direct measurement of the anterior rectal wall dose for
each treatment fraction, increasing the accuracy of toxicity correlation.
In the case of variation in balloon positioning between fractions, multiple dual
MOSkin detectors could be placed in a strip along the superior-inferior axis of the
balloon. This would provide dose measurement at multiple locations along the anterior
rectal wall relative to the PTV and improve the probability of measurement of the
highest anterior rectal wall dose.
The MOSkin was able to be read out at 1Hz during the delivery of a radiotherapy
treatment fraction. The ability to read out the MOSkin in real-time allows the rectal
7.5. Conclusion
155
dose to be tracked as it is being delivered, potentially providing a 'dose alarm' if the
rectal wall dose exceeds a set tolerance.
7.5 Conclusion
A novel real time in vivo dosimetry system for tracking of rectal wall doses in prostate
radiotherapy has been presented. The dosimetry system provides accurate, real time
tracking of the rectal wall dose during the delivery of a treatment fraction. The
MOSkin showed angular response leading to a lower measured dose than that of EBT
lm. Two correction methods were presented to negate the eects of the angular
response. The rst correction method, application of a copper ltering layer at the
top of the MOSkin detector reduced the dierence between 'face up' and 'face down'
MOSkin responses. The second correction method, the dual MOSkin conguration,
resulted in a reduction of the angular response to within 2%. The dual MOSkin
was then used to verify the anterior rectal wall dose for a 3DCRT and IMRT plan for
a hypothetical prostate target in a phantom. The dual MOSkin provided real time
measurement of the anterior rectal wall dose, with measurements 2.62% (3DCRT)
and 3.17% (IMRT) lower than the Pinnacle RTPS prediction. It is expected that
the dual MOSkin correction method will be pursued in future rectal balloon in vivo
measurements.
Chapter 8
Novel surface detectors applied to
total scalp irradiation with helical
tomotherapy
8.1 Introduction
Irradiation of the total scalp is a treatment technique used for a variety of supercial malignancies. Traditional techniques for total scalp irradiation include use of a
combination of electron and photon beams (Akazawa, 1989) and more recently, linac
based intensity modulation radiation therapy (IMRT) (Bedford et al. , 2005) and serial
Tomotherapy (Locke et al. , 2002). Helical tomotherapy, an image guided intensity
modulation therapy system, has been shown to be an eective means of delivering
total scalp radiotherapy (Orton et al. , 2005). The advantages of using helical tomotherapy over conventional Linac based techniques for total scalp irradiation have
Part
of this chapter has been published in Medical Physics:
Hardcastle N, Soisson E, Metcalfe P E, Rosenfeld A
B,
Tom
e W A, 2008,
Medical
Physics,
Dosimetric Verication of Total Scalp Irradiation with Helical Tomotherapy,
volume 35, issue 11, pages 5061-8
156
8.1. Introduction
157
been reported (Khuntia et al. , 2006; Orton et al. , 2005). These advantages include
alleviating the need for complicated beam matching such as that required for combination electron-photon treatments and the ability to achieve uniform coverage of the
PTV without increasing dose to normal brain as compared to other types of IMRT delivery (Khuntia et al. , 2006). This superior target coverage can often be accomplished
with tomotherapy since the delivery uses a larger number of beam angles, allowing
for more conformal treatment of concave structures. Tomotherapy's 360 delivery is
separated into 51 projections per rotation. Due to the machine geometry, the binary
MLC divides the beam into beamlets that can be tangential to the scalp at any given
projection, increasing delivered supercial dose. Directional blocking is employed on
the brain which eectively forces the majority of the beamlets to be delivered to the
scalp PTV tangentially, since beamlets cannot enter through the brain to deposit dose
in the PTV but only exit through brain after depositing dose in the PTV. The helical
tomotherapy delivery results in a very ne and customisable modulation resolution.
As the scalp diameter changes, tomotherapy's binary MLC allows fast adaptation of
the eld width and position. In addition, an on-board CT detector allows for image
guidance through the acquisition of pre-treatment Megavoltage CT (MVCT) images
prior to each treatment fraction, increasing setup reproducibility and accuracy (Li
et al. , 2007).
A potential concern when using helical tomotherapy for total scalp irradiation
has recently arisen, with recent published work showing that the TomoTherapy (TomoTherapy Inc. Madison, WI) planning system can over estimate the calculated
supercial dose for head and neck treatments. Ramsey et al. (2007) found that the
supercial dose (dose over rst 2mm depth) was over estimated by 3-13% for a parotidsparing IMRT treatment of head and neck cancer. Higgins et al. (2007) showed that
the calculated supercial dose was 10% greater than the measured dose for a typ-
8.1. Introduction
158
ical oropharynx treatment at lateral locations yet was accurate at anterior locations.
Both of these studies used a combination of Thermoluminescent Dosimeters (TLDs)
and radiographic lm for dose verication in an anthropomorphic phantom. A study
by Cheek et al. (2007) found that for supercial PTVs between 2 and 6 cm deep in
a polystyrene phantom the TomoTherapy dose calculation over-estimated the radiographic lm measured dose in the rst 10mm depth by up to 8%.
As mentioned above, total scalp irradiation with helical tomotherapy includes a
large number of beamlets that are delivered primarily tangential to the surface. This
diers from the above-mentioned head and neck treatment where the beamlets are
predominantly orthogonal to the patient surface. If the TomoTherapy planning system is over-estimating the delivered supercial dose as shown in the above-mentioned
reports, then a total-scalp PTV will be under-dosed. A bolus material may be required
to increase the supercial dose if the target volume extends to the surface, such as used
with other treatment techniques. With the tangential beam arrangement it is possible
that the calculated dose may be accurate, and bolus material may not be required to
achieve adequate target coverage.
In this study, the supercial dose was measured for a total scalp treatment delivered
to an anthropomorphic head phantom with helical tomotherapy. Radiochromic and
radiographic lm as well as the MOSkin detection system was used to compare the
measured supercial dose with the calculated dose from the TomoTherapy treatment
planning system.
159
8.2. Method
Table 8.1: Optimization parameters for helical tomotherapy total scalp treatment
ROI Weight Max Max dose DVH DVH Min Min dose DVH
dose penalty vol(%) dose dose penalty penalty
PTV 100 40
300
98
40 40
100
Brain 20
38
75
10 2.5 75
8.2 Method
8.2.1 Treatment plan
A treatment planning CT was taken of an anthropomorphic head phantom (RANDO,
The Phantom Laboratory, Salem, NY). The image set was transferred to the Pinnacle
treatment planning system (Philips Radiation Oncology Systems, Fitchburg, WI) and
a hypothetical target volume was contoured to simulate a typical total scalp PTV.
The contours were based on a patient treated at this institution. The brain was
contoured as a critical structure. The CT data and the contours were then transferred
to the TomoTherapy treatment planning system (v. 2.2.4.1.1. TomoTherapy Inc.
Madison, WI). A plan was generated with 40Gy prescribed to 98% of the PTV using
the optimization parameters given in Table 1. Directional blocking was used for the
brain structure. Directional blocking dictates that individual beamlets cannot enter
through the brain to deposit dose in the PTV, but can exit through the brain after
depositing dose in the PTV. A eld width of 2.5cm, a pitch of 0.3 and a (maximum)
modulation factor of 3 were used. The ne calculation grid was employed meaning
that the dose grid matched the 256x256 downsampled CT, giving a dose grid resolution
of 0.1875 x 0.1875 x 2.5 mm . A custom Aquaplast (Uni-Frame, CIVCO, Kalona,
Iowa) head mask was created to assist in phantom setup. The resultant dose-volume
histogram (DVH) and calculated dose distributions are given in Figure 8.1.
For all measurements a Mega-Voltage CT (MVCT) scan was taken prior to delivery.
A 'normal' slice resolution (pitch = 2) was used giving a slice width of 4mm. This was
3
8.2. Method
160
Figure 8.1: Resultant dose distributions and cumulative dose volume histogram for
scalp treatment
then aligned to the planning CT image. The accuracy to which the MVCT can be
registered to the planning CT has been measured previously for this anthropomorphic
head phantom (Boswell et al. , 2006) and was found to be within 1mm. The dose
delivered to the MVCT has been measured to be less than 1.5cGy for this pitch which
constitutes less than 1% of the prescription dose for this treatment (Shah et al. , 2008).
8.2.2 Transverse measurements
Gafchromic EBT radiochromic lm sheets (ISP Corp, Wayne, NJ, lot #: 47207-02I)
were cut to the shape of the phantom and placed in the transverse slices of the phantom. Handling and storage protocols were followed (Niroomand-Rad et al. , 1998).
The lm was cut so that it extended < 4mm outside of the phantom rather than to
match the phantom edge so as to negate eects of polymer damage on the lm edge
due to cutting (Yu et al. , 2006).
The EBT lm was scanned on an Epson Perfection V700 atbed scanner. All lms
were scanned at least 24 hour post irradiation to minimise post-irradiation color eects
161
8.2. Method
(Cheung et al. , 2005). The scanning resolution was 75dpi and all lms were scanned in
the landscape orientation. The resultant images were 32-bit color with the red channel
chosen for analysis. All analysis was done on a desktop PC using ImageJ and Matlab
software. Background corrections were made to remove spatial non-uniformities by
scanning each EBT lm prior to irradiation. Calibration was performed by plotting a
5th order polynomial curve to nine dose points from 0 - 430cGy that were measured
on separate lms using the TomoTherapy beam. The EBT lm was calibrated in the
Tomotherapy beam in the same method presented in Chapter 6. A maximum error of
0.6% existed between the polynomial and the measured dose.
The transverse dose distributions were also measured using Kodak EDR2 lm
(Kodak, Rochester, NY). The lm was cut to the shape of the phantom slices and black
insulation tape was used to light-seal the lm edges. The combination of minimizing
the eect of polymer damage on the lm edge and reducing the amount of tape inside
the phantom (to eliminate air gaps) meant the lm extended < 12mm outside of the
phantom. Film calibration was performed using the same procedure as for the EBT
lm. A Vidar (VXR-16 Dosimetry Pro, Vidar Systems Corp, Herndon, VA) scanner
was used to scan all of the lms. A resolution of 71dpi was used. The resultant images
were 16-bit. Analysis was performed using ImageJ software. Investigation was done
to determine whether lm extending out of the phantom changed the build up dose
on the surface. Both types of lm were irradiated edge-on whilst sandwiched between
solid water slabs. A 10x10cm 6MV linac eld was used. The length of lm hanging
outside of the solid water was varied from 0 to 8mm in 4mm increments.
2
8.2.3 Surface measurements
Sheets of EBT lm (1.5 x 6 cm ) were cut and placed on the surface of the phantom
at selected locations. The suitability of EBT lm for surface dosimetry has previously
2
162
8.2. Method
been reported (Devic et al. , 2006).The surface dose measured by the EBT lm is
equivalent to the dose measured at an eective depth of 153m in water. EBT lm
measurements were repeated twice.
The surface dose was also measured using the MOSkin. This is a research MOSFET
detector designed specically for skin dosimetry by the Centre for Medical Radiation
Physics (CMRP), University of Wollongong, Australia. MOSFET detectors have been
shown to be eective skin dosimeters due to their small water equivalent depth (WED)
(Butson et al. , 1996; Scalchi et al. , 2005; Cherpak et al. , 2007; Xiang et al. , 2007).
Butson et al. (1996) demonstrated that the MOSFET detector without packaging
can measure surface dose in good agreement with an ATTIX parallel plate ionization
chamber. For practical applications the MOSFET chip must be protected from the
environment. Usually this is achieved by covering the MOSFET chip with an epoxy
bubble. However, it is dicult to reproduce the epoxy bubble dimensions. This leads
to variations in the WED ranging from 0.7-1.8mm (Butson et al. , 1996; Scalchi et al.
, 2005; Cherpak et al. , 2007). The MOSkin detector design aims to alleviate the
non-uniform WED from the epoxy bubble, yet still retain protection from the environment. This was achieved by using exible carrier based on 20m thick polyamide
lm. Use of the polyamide lm gives a reproducible uniform build-up designed for
surface dosimetry, with a WED of 70m or other as required (the ICRP recommended
skin depth for radiation skin damage is 70m (ICRP, 1991)). The size of the MOSFET silicon chip is 0.6x0.8x0.35mm . The polyamide lm on the top of the MOSFET
chip serves as the MOSFET carrier and provides reproducible build-up, protection of
the surface of the MOSFET chip from environment conditions and connection of the
MOSFET to the reader. For hermetic sealing the MOSFET, with attached polyamide
carrier above the MOSFET chip, is placed in a plastic package with an opening of
the same size as a MOSFET chip. This is then sealed from the back side. This de3
163
8.2. Method
sign avoids wire bonding of the chip and metal connection pads close to the sensitive
volume of the MOSFET. The total detector size is 2x5x0.7mm . MOSkin detectors
were calibrated in a conventional linear accelerator in a 6MV beam then placed at
four dierent locations on the phantom surface that were also measured with EBT
lm. The change in threshold voltage of the MOSkin post-irradiation was read out
using a CMRP developed reader. The MOSkin measurements were repeated twice for
statistical accuracy.
The WED of measurement was veried for the MOSkin and the EBT lm using
Monte Carlo (MC) and Attix chamber data. Each detector (MOSkin, EBT lm and
Attix chamber), in turn, was placed on the surface of a slab phantom and irradiated
with 10x10cm and 2.5x2.5cm 6MV elds from a Varian 21EX linac. Each detector
was then placed at Dmax (1.5cm) and irradiated with the same dose as that for the surface measurements. The surface dose was then normalised to the Dmax dose. No Attix
chamber measurement was performed for the 2.5x2.5cm eld as this eld's dimensions
were smaller than that of the detector. The measured surface doses were compared
with MC simulation data. Two MC codes were used - BEAMnrc and Geant4. Detailed
information and commissioning plots for the BEAMnrc model is given in Appendix B.
The BEAMnrc simulation was performed in two steps. The rst step was the creation
of a phase space le at 100cm from the target. This step involved detailed simulation of a Varian 21EX linac treatment head. A 6.3MeV (0.2MeV FWHM) electron
beam (50 x 10 electrons) was incident on a Tungsten target. Energy cut o values
of 0.521MeV and 0.01MeV for electrons and photons respectively were used. These
correspond to the lowest energies for which data exists in the PEGS4 material. It is
important for these values to be as low as possible for surface dose measurement as
the dose to the surface is due to low energy photons and electron contamination. The
resultant phase space le was then used as the input for a DOSXYZnrc simulation.
3
2
2
2
6
164
8.2. Method
Table 8.2: Example of MOSkin data collection spreadsheet. V is the initial threshold
voltage, V is the threshold voltage 30s post-irradiation, and V is the change in
threshold voltage.
Angle MU V (V) V (V) V (V) % of Dmax
0 200 9.259 9.355 0.096
17.52
15 200 9.448 9.546 0.098
17.88
30 200 9.543 9.646 0.103
18.80
45 200 9.643 9.763 0.120
21.90
60 200 9.759 9.902 0.143
26.09
75 200 9.897 10.089 0.192
35.04
0
0
The geometry of the phantom was a 30x30x30cm cube water phantom. The water
phantom was split into a number of voxels. The resolution of the voxels in the x and
y directions was 2x2cm . In the depth direction the resolution was 100m for the rst
1.5cm with 100m thick voxels at 5cm and 10cm depth. The Geant4 simulation data
was provided by a fellow CMRP PhD Candidate (Oborn, 2008). The water phantom
dimensions and voxel resolution was the same as that for the DOSXYZnrc simulation.
The electron and photon energy cut o values were 0.521MeV and 0.01MeV respectively. The measured data, compared with the MC data is given in Figure 8.2, which
shows that the EBT Film and MOSkin surface doses agree well with two MC code
calculations of the surface dose at their theoretical WED for both eld sizes.
The response of the EBT Film and MOSkin detectors to oblique beam incidence
was also investigated. In separate measurements, the two detectors were placed on
the surface of a 30x30x30cm cube water phantom. Using a Varian 21EX linac 6MV
photon beam and 10x10cm and 2.5x2.5cm eld sizes, the detectors were irradiated
from gantry angles of 0 to 85 in 15 increments. The measured surface doses at
each angle were normalised to the dose at Dmax (1.5cm) for each eld size. The data
was calculated in a spreadsheet during collection. An example of the collected data is
presented in Table 8.2. The results are given in Figure 8.3, which show that the two
detector response diers as a function of detector WED and eld size. The surface
3
2
3
2
2
8.3. Results and discussion
165
dose measured with EBT Film is always higher than that with the MOSkin, however
the ratio of the EBT Film to MOSkin changes with angle and eld size.
8.3 Results and discussion
8.3.1 Transverse measurements
The EBT and EDR2 transverse lm images were converted to absolute dose and scaled
to show the total dose over the 20 fraction treatment. The locations of the lms and
resultant EBT lm measured dose maps are given in Figure 8.4.
Line proles were taken across the lms and compared with line proles taken from
the calculated dose grid. The phantom has support pegs; the holes for these were used
to align the dose proles. Proles were taken 2.5cm posterior to the peg holes on the
digitised lm images. The plan proles were taken from the calculated dose cube.
The calculated dose grid slices were not parallel to the lm slices. However, a smaller
region of each slice in the dose grid corresponded to the same spatial region in the
lm slice. The proles from the calculated dose grid were then taken from the slice
containing the region where the lm prole was taken. A line was drawn on the lm
using a scalpel at the phantom edge to determine the location of the surface of the
phantom. This was visible on the digitised lm images and had a width of 0.034cm
(one pixel width) at the location of the proles. No distortion of the dose either side
of the scalpel line was observed. The pixel immediately medial to the line was taken
as the start of the phantom. The line did not distort the dose values either side of it.
For the dose grid, the rst voxel with a Hounseld Unit (HU) between that of water
and air was taken as the start of the phantom. Proles were taken in regions where
the immobilizing head mask was not in contact with the phantom surface so the start
of the phantom was easily dierentiated.
166
8.3. Results and discussion
Figure 8.2: (a) 10x10cm and (b)2.5x2.5cm eld depth dose curves with MOSkin,
EBT Film and Attix chamber surface measurements compared with BEAMnrc and
Geant4 (Geant4 data courtesy of Oborn (2008), private communication) MC simulation data. The depth axis is displayed on a logarithmic scale to show the detail of the
buildup.
2
2
167
8.3. Results and discussion
Figure 8.3: Surface dose measurements as a function of incident beam angle for (a)
10x10cm eld and (b) 2.5x2.5cm eld. The ratio of the EBT lm to the MOSkin
measurement changes based on angle and eld size.
2
2
8.3. Results and discussion
168
Figure 8.4: Transverse lm locations and resultant digitised lm images. The black
dotted line shows the location of the phantom edge. The black lines on sheets 1 and
2 show the locations of the proles shown in Figures 8.6 and 8.7
8.3. Results and discussion
169
Figure 8.5: Buildup curves for (a) EBT lm and (b) EDR2 lm as a function of length
of lm protruding out of solid water slabs and irradiated edge on parallel to 6MV
photon beam central axis
170
8.3. Results and discussion
The eect of edge-on irradiation of lm protruding out of the phantom was investigated. These results are shown in Figure 8.5. Each prole is normalised to the dose
at 1.5cm depth. It was found that for the EBT lm the surface dose measurement
was not aected by lm protruding out of the phantom. For EDR2 lm it was seen
that an over response at the surface followed by an under response in the build up
occurred with the lm extending out of the phantom. This is probably due to the
higher eective atomic number of the EDR2 and is a similar result to that found by
Ramsey et al. (2007).
The measured proles compared with the calculated proles for EBT Film and
EDR2 Film are shown in Figures 8.6 and 8.7 respectively. The dose calculation, when
performed on the smallest dose grid resolution, obtains the dose in a 1.875x1.875x2.5mm
sized voxel. The digitised EBT lm images have a pixel size of 0.34x0.34mm and the
digitised EDR2 lm images have a pixel size of 0.36x0.36mm . The proles show that
the calculated build up dose agrees with the measured EBT lm dose. In Figure 8.6(a)
the EBT lm measured supercial dose was seen to increase from 33.6Gy to 41.2Gy
over the rst 2mm of depth in the phantom. The corresponding calculated dose increased from 32.8Gy to 40.8Gy, a dierence in surface dose of 2.4% of the measured
dose. In Figure 8.6(b) the EBT lm measured supercial dose was seen to increase from
36.5Gy to 41.9Gy over the rst 2mm. The corresponding calculated dose increased
from 36.8Gy to 42.4Gy, a dierence in surface dose of 0.8%. Figure 8.6(c) shows in
this location, the EBT lm measured surface dose was 43.4Gy. The corresponding
calculated dose surface dose was 42.8Gy giving a dierence between measured and
calculated dose of 1.5%. No build up of dose was observed in this location due to the
bolus eect of the head rest used to aid treatment setup.
Figure 8.7 (a) and Figure 8.7 (b) also show that the EDR2 lm surface dose is
greater than the EBT lm and calculation. It then increases at a slower rate than
2
2
3
8.3. Results and discussion
171
Figure 8.6: (a) Cross plane prole of transverse sheet 1 taken 1cm under peg holes
for EBT lm and plan data. (b) Cross plane prole of transverse sheet 2 taken 2.5cm
under peg holes for EBT lm and plan data. Zoomed in section shows rst 1cm depth
in phantom. (c) Posterior-Anterior prole taken across transverse sheet 2 along the
centre of the lm for EBT lm and plan data. The locations of the proles are shown
in Figure 8.4
8.3. Results and discussion
Figure 8.7: The same proles as in 8.6 but with EDR2 data
172
8.3. Results and discussion
173
Figure 8.8: Surface EBT lm locations and measured doses.
that of the EBT lm and calculation. In Figure 8.7(a) the EDR2 lm measured dose
increases from 36.2Gy to 38.3Gy over the rst 2mm depth in the phantom. In Figure
8.7 (b) the EDR2 lm measured dose increases from 38.1Gy to 41.4Gy over the rst
2mm depth in the phantom. For depths in the phantom beyond approximately 5mm
the EDR2 lm dose agrees with the calculated and the EBT lm dose.
8.3. Results and discussion
174
Figure 8.9: Comparison of MOSkin measured dose and EBT lm surface dose.
8.3.2 Surface measurements
Films were cut and placed on the surface of the phantom in various locations. The lms
were converted to dose maps and scaled to show the total dose over 20 fractions. The
location and doses measured are given in Figure 8.8. Two features are immediately
visible. The rst is the eect of the head mask on the surface dose. The grid pattern
seen under the head mask shows a dose dierential between regions immediately under
the head mask and adjacent regions of up to 15Gy over the total treatment. The head
mask is acting eectively like a bolus material in the regions where it is in direct contact
with the skin. The second is the eect of the head rest on the surface dose. The head
rest provides sucient build up to obtain the prescription dose on the surface.
MOSkin detectors were placed at four locations on the head phantom that corresponded to locations that surface EBT lms were also placed. The MOSkin-measured
doses as compared with the EBT lm surface dose measurements are given in Figure
8.9. The MOSkin-measured doses are less than that measured with the EBT lm.
This is due to the dierence in WED of measurement between the MOSkin detectors
and EBT lm, shown in Figure 8.2. The eective depth of measurement in water for
8.3. Results and discussion
175
the MOSkin is 70m where as for the EBT lm it is 153m. At these depths there
exists large charged-particle disequilibrium. Any small change in depth will lead to
a large change in the measured dose. The consequence of the dierence in WED is
that the MOSkin will measure a lower dose than the EBT lm. The magnitude of this
dierence will vary with angle of incidence (as shown in Figure 8.3); the WED of the
two detectors will increase and represent dierent depths on the build-up curve. The
ratio of the two measured doses diers according to the change in gradient of the dose
build-up. Additionally, the MOSkin has a much smaller sensitive depth over which
dose is measured (0.55m) compared with the EBT lm (40m). This is another factor that can be taken into consideration when comparing doses in a steep dose gradient
such as that on the surface.
The treatment plan delivers the radiation dose with beamlets that are primarily
tangential to the scalp. This is a direct consequence of the directional blocking applied to the brain contour. Beamlets are not allowed to pass through the brain to
deliver dose to the PTV therefore they must be delivered tangentially to satisfy the
DVH constraints and the directional block. Figure 8.10 shows a sample (every third
projection angle) of the planned leaf opening times. It is clear from this gure that the
majority of the beamlets were delivered tangentially to the phantom surface. Beamlets
that are tangential to the surface increase the surface dose (Lee et al. , 2002; Chow &
Grigorov, 2008). The TomoTherapy planning system accurately calculated the surface
dose for this particular treatment. This is in contrast to previous reports regarding
the TomoTherapy planning system surface dose calculations. Previous reports employ
a larger dose calculation grid which would lead to a higher surface dose calculation.
These reports also do not utilise directional blocking and are targeting deep seated
PTVs, thus it is expected that tangential beamlets would not make up a large proportion of the delivered sinogram. The combination of these two dierences most
8.3. Results and discussion
176
Figure 8.10: Sample (every third projection shown) of the incident uence sinogram
for one rotation in the centre (superior-inferior direction) of the PTV. On each chart
the abscissa axis is MLC leaf number and the ordinate axis is relative planned leaf
opening times. The MLC predominantly blocks the central beamlets of the fan beam
and allows beamlets through that are tangential to the scalp.
8.4. Conclusion
177
probably explains the contradiction between this report and previous studies. For this
reason, it is suggested that accurate calculation of the dose in electronic disequilibrium
regions requires a dose grid as ne as possible.
If required, a bolus layer may be employed to increase the surface dose to the
prescription dose, as discussed in a recent report (Lin et al. , 2008). Another option to
achieve a more homogeneous surface dose is to mould a custom thermoplastic helmet
for each patient. This would bring the supercial dose up to the prescription dose
with possibly the same level of setup uncertainty as a thermoplastic immobilization
head mask.
8.4 Conclusion
The utility of Gafchromic EBT lm and the MOSkin for surface dosimetry was investigated. EBT Film and the MOSkin were found to provide high resolution surface
dose measurements at WEDs of 153m and 70m respectively. The EBT Film and
MOSkin detectors were then used with Kodak EDR2 Film to verify the supercial
dose for a total scalp irradiation using helical Tomotherapy.
The Gafchromic EBT lm was found to be an excellent dosimeter for this application. Its relatively low eective atomic number (Z=6.9) is close to that of water thus
build up eects with edge-on irradiation is not an issue as with EDR2 lm, where the
high atomic number and packaging of the lm aect the dose measurement. When
compared with EBT lm surface measurements the MOSkin measured a lower surface
dose due to the shallower WED of the detection volume.
Calculated surface doses using the TomoTherapy RTPS were not able to be compared with surface EBT lms or MOSkin measurements as they are essentially measuring the dose at dierent locations. The surface doses calculated by the Tomotherapy
planning system agree with that measured with EBT lm for a total scalp irradiation
8.4. Conclusion
178
to within 2.5% of the measured dose on the transverse EBT lms.These ndings are
in contrast to previous reports in the literature on Tomotherapy surface dose calculations, which suggested the TomoTherapy planning system over-estimated the surface
dose (Ramsey et al. , 2007; Higgins et al. , 2007; Cheek et al. , 2007). This is most
likely a consequence of the combination of a smaller dose grid resolution utilised and
this particular beam arrangement, which includes primarily tangential beamlets as opposed to beamlets orthogonal to the patient surface as in conventional head and neck
treatments. When the doses from multiple tangential beamlets are super-imposed the
depth required for full dose build up is reduced and a greater surface dose is achieved.
Chapter 9
Multileaf collimator end leaf
leakage: Implications for wide-eld
IMRT
9.1 Introduction
9.1.1 MLC leaves and carriages
The leaves of the Millennium MLC (Varian Medical Systems, Palo Alto, CA, USA) are
mounted on movable carriages and each leaf can extend 14.5cm from the carriage. To
limit radiation damage of the electronic components the carriages must be shielded by
the jaws at all times. Intensity modulated radiotherapy (IMRT) can involve treatment
volumes that have cross-sectional dimensions greater than 14.5cm. To perform intensity modulation of a eld wider than 14.5cm using the Millennium MLC the ends of
closed leaf pairs must be positioned inside the eld. The Millennium MLC is designed
Part
of this chapter has been published in Physics in Medicine and Biology:
Hardcastle N, Metcalfe P E, Ceylan A, Williams M J, 2007, Multileaf Collimator End Leaf Leakage:
Implications for wide-eld IMRT, Physics in Medicine and Biology, volume 52, issue 21, pages N493N504
179
9.1. Introduction
180
with rounded leaf ends to approximate focusing in the direction of leaf motion. A
consequence of the rounded design is the partial transmission of radiation through the
leaf ends, which results in additional dose to the patient and this occurs even when
the leaves are completely closed. The dosimetric implications of the rounded leaf end
design for IMRT have been widely investigated (Boyer & Li, 1997; LoSasso et al. ,
1998; Arneld et al. , 2000b) and a common method employed to account for the
partial transmission is to apply an inward shift to the leaf positions. The inward shift,
referred to as the radiation eld oset (RFO) is dened as the dierence between the
width of the radiation eld in the direction of leaf travel and the width of the light eld
in the direction of leaf travel. The RFO approximates the increase in the penumbra
width caused by the transmission through the leaf end and is reported to be between
0.2-0.3mm for a Varian MLC when using 6MV photons (Kung & Chen, 2000). The
RFO method is adequate for improving the agreement between planned and delivered
doses for open MLC apertures. The RFO method is not applicable to closed leaf pairs
as the leaves cannot be physically moved closer together. The radiation transmitted
through the closed leaves of a Millennium MLC has been reported to be 25.3% for
6MV photons (Heath & Seuntjens, 2003). The limited leaf extension, a restriction on
carriage movement during delivery and end leaf leakage has inhibited the use of the
Varian MLC for treating IMRT elds wider than 14.5cm.
9.1.2 Wide eld IMRT with the Varian Millenium MLC
MacKenzie et al. (2002) proposed three methods to achieve an IMRT delivery using
the Varian Millenium MLC for volumes wider than 14.5cm. One method was to restrict
the eld width to 14.5cm and use collimator rotation to attempt to cover the volume.
A second approach was to use the wider eld but apply static feathering of the closed
leaves, such that the closed leaf pairs were stepped across the eld in the direction of
9.1. Introduction
181
Figure 9.1: Schematic showing head and neck IMRT treatment using (a) split coaxial
overlapped elds and (b) a single wide eld
leaf travel as a function of the monitor units (MUs). The static feathering distributed
the end leaf leakage across the eld thereby reducing the dose due to leakage at any
one particular point. The third approach was to deliver two smaller overlapping elds
from the same gantry angle; this technique has been investigated by others Wu et al.
(2000); Dogan et al. (2003); Metcalfe et al. (2004). An example of the split co-axial
elds is illustrated in Figure 9.1. This technique avoids the problem of end leaf leakage
however it does require more elds to perform the IMRT delivery. Field matching can
also become an issue, as day to day positional variations can aect eld overlap regions.
The methods discussed previously have used either a spatial approximation for the
transmission through the leaf ends, or attempted to minimise the eects of end leaf
leakage by distributing closed leaf pairs across or outside the eld. An alternative
method is to incorporate the rounded leaf end design into the radiotherapy treatment planning system and intrinsically account for the transmission during the plan
construction (Cadman, McNutt et al. 2005; Williams and Metcalfe 2006).
9.1. Introduction
182
9.1.3 Wide eld IMRT in the Pinnacle RTPS
The Pinnacle radiotherapy treatment planning system (RTPS), version 7.4 and higher
(Philips Radiation Oncology Systems, Milpitas, CA, USA) provides a rounded leaf
end MLC model. A detailed description of the model is provided by Cadman et al.
(2005). Part of the increased MLC functionality has been the inclusion of an option
for delivering step-and-shoot IMRT with a single wide eld when using the Varian
MLC; this is in addition to the only option available previously of using split elds.
Using this option, for elds wider than 14.5cm closed leaf pairs will occur in the eld.
Due to a 14.5cm limit on over-travel of the leaves, the location of closed leaf pairs is
restricted to a 29cm band in the centre of the eld. By applying a similar approach
to that proposed by MacKenzie et al. (2002) the closed leaf pairs are distributed
throughout this central 29cm band. The process of distributing the closed leaves has
been illustrated in Figure 9.2; the position of each closed leaf pair is dependent on the
positions of the nearest leaf openings and on the leaf-extension limit. As the segment
shapes change the position of the closed leaves change, and as the eld width increases
the range of possible positions for the closed leaves decreases. This has an upper limit
of 29cm where the only location for closed leaf pairs is along the centre line of the
eld. Unlike the method of MacKenzie et al. (2002) in which the closed leaves were
evenly distributed across the central 29cm of the eld, this method relies on the leaf
positions of adjacent segments. There is the possibility that a high concentration of
closed leaf pairs could occur in the one area due to multiple segments in an intensity
modulated beam having very similar shapes.
In this study the end leaf leakage of a Millennium MLC using radiographic and
radiochromic lm for 6MV photons was characterised. The accuracy of the Pinnacle
RTPS in modelling the end leaf leakage is veried for both single segments and for an
IMRT eld, and the implications of end leaf leakage for wide eld IMRT are discussed.
183
9.2. Method
Figure 9.2: Wide eld IMRT as applied with the Pinnacle RTPS. All closed leaf pairs
above the topmost section are positioned at the midpoint of the topmost leaf opening
and all closed leaf pairs below the lowermost section are positioned at the midpoint
of the lowermost leaf opening. Closed leaf pairs that occur between two openings are
positioned at the average of the midpoints of the two nearest leaf openings
9.2 Method
All measurements were performed using 6MV photons generated on a Varian 21EX
linear accelerator equipped with a Millennium 120 leaf MLC. All exposures were made
using 300 MUs unless stated otherwise. The MLC apertures were created with the
Varian MLC Shaper software (V.6.1), the IMRT eld and the planar dose distributions
were generated with Pinnacle (V.7.6c).
9.2.1 Magnitude of end leaf leakage
The dependence of the end leaf leakage on the gap width and o-axis position has
been measured. Symmetric MLC gap widths of 0.0, 0.6, 1.0, 2.0, 3.0, 5.0 and 10.0mm
were created and the secondary jaws used to dene a 10x10cm eld above the MLC.
The mechanical calibration of the MLC had been performed to within a precision of
2
184
9.2. Method
0.05mm. The MLC control system on the linear accelerator forces a minimum leaf
separation of 0.5mm for moving leaves, therefore to ensure that all elds generated by
the RTPS could be physically delivered the RTPS had to incorporate this minimum
separation. To eliminate the possibility of rounding errors a minimum leaf separation
of 0.6mm was specied for the Pinnacle MLC model used in this study. Any closed
leaf pairs that occur in a Pinnacle generated plan will be at the minimum separation
of 0.6mm, hence this gap width was included in the measurements. The specic values
for the parameters used in the Pinnacle MLC model have been described previously
(Williams & Metcalfe, 2006).
O-axis leakage measurements were also performed for the 0.0, 0.6 and 3.0mm gap
widths. Symmetric and asymmetric 20x20cm elds, as dened by the jaws, were used
in conjunction with the MLC gaps positioned at o-axis distances of -5, 0, 5, 10 and
15cm. All leakage measurements were performed with radiographic and radiochromic
lm.
2
9.2.1.1 Film measurements
Sheets of EDR2 radiographic lm (Eastman Kodak Company, Rochester, NY, USA)
were placed at a depth of 1.5cm (dmax for 6MV) in solid water (RMI-457, GammexRMI, Middelton, WI, USA). The surface of the solid water phantom was positioned at
100cm source to surface distance. The developed lms were scanned on a VXR 12-bit
lm scanner (VIDAR Systems Corporation, Herndon, VA, USA) at a resolution of 300
dpi. The conversion of optical density to dose was based on a third order polynomial
t of the data for a set of 16 calibration lms taken over the range of 0Gy to 3.5Gy in
25cGy steps (Williamson, Khan et al. 1981; Suchowerska, Davison et al. 1997). The
EDR2 measurements were repeated with Gafchromic EBT lm (ISP Corp, Wayne,
NJ, USA) under identical set-up conditions. The EBT lms were scanned on a Umax
Astra 6700 (Umax Technologies, Inc., Taiwan) atbed scanner in reective mode at a
9.2. Method
185
Figure 9.3: The end leaf leakage for a 6MV photon beam measured at a depth of 1.5cm
in solid water using EDR2 lm for (a) 0mm gap width (b) and 3mm gap width
resolution of 300 dpi. A conversion from grey scale to dose was carried out using a set
of calibration lms performed at the same dose intervals as the EDR2 calibration.
9.2.1.2 Pinnacle dose maps
Each of the eld arrangements were reproduced on Pinnacle and a 2D planar dose
distribution calculated at a depth of 1.5cm in water. For the 0.0mm gap width the
minimum leaf separation parameter on Pinnacle was reduced to 0mm, for all other
elds it remained at 0.6mm. Dose calculations were performed using the collapsed cone
convolution algorithm with a 2mm dose grid and a matching uence grid resolution.
Dose proles across the leaf ends were extracted from the 2D dose distributions and
compared to the dose proles measured with lm.
9.2.2 IMRT eld
Film dosimetry was performed for an IMRT eld that had been used clinically for
the treatment of a head and neck case. This particular IMRT case involved 7 beams
9.3. Results and discussion
186
delivering 50Gy in 25 fractions. An IMRT eld that exhibited end leaf leakage was
chosen from the 7 beams. This IMRT eld was 18cm wide and required 21 segments
with 130MU in total. The planar dose distribution at a depth of 10cm (100cm SAD) in
solid water was measured with the EDR2 and EBT lm. The dose distributions were
scaled by a factor of 25 to reect the total dose for the full course of treatment, and
the lm measurements were compared to the dose distribution predicted by Pinnacle.
9.3 Results and discussion
9.3.1 Magnitude of end leaf leakage
9.3.1.1 Central axis end leaf leakage
In Figure 9.3 the EDR2 lms exposed at a depth of 1.5cm in solid water for the
0mm and 3mm gap widths are shown. The corresponding dose proles across the leaf
ends measured with the EDR2 and EBT lm and predicted by Pinnacle are shown in
Figure 9.4; the data for the 0.6mm gap has also been included. For the closed MLC, or
0mm gap width, the maximum leakage measured with the EBT lm was 0.39cGy/MU.
This value was under-predicted by Pinnacle by approximately 40%, which calculated
a leakage of 0.23cGy/MU. The maximum leakage measured for the 0.6mm gap was
0.51cGy/MU, which is approximately half of the dose received by a point in the open
eld. As mentioned previously the 0.6mm gap width represents the spacing between
closed leaf pairs for clinical IMRT elds generated using Pinnacle.
The peak measured and Pinnacle predicted doses for all gap widths are shown
in Figure 9.5. For gap widths less than 5mm Pinnacle under-estimated the dose
between the closed leaf pairs. Investigation of the dose grid resolution showed that
there was no change in the dose predicted by Pinnacle for grid sizes of 2mm and
smaller. While the grid size used was larger than the smallest leaf gap, it was smaller
9.3. Results and discussion
187
Figure 9.4: Line proles across the end leaf leakage for a 6MV photon beam measured
at a depth of 1.5cm in solid water with EDR2 and EBT lm, and predicted by Pinnacle
for the (a) 0mm (b) 0.6mm and (c) 3mm gap widths.
9.3. Results and discussion
188
Figure 9.5: Comparison of a) the Pinnacle predicted and measured doses for the end
leaf leakage and b) FWHM of end leaf leakage peaks as a function of width between
opposing MLC leaves
9.3. Results and discussion
189
than the FWHM of the leakage dose as measured using EDR2 and EBT lm. At
a grid size of 4mm Pinnacle underestimated the dose by approximately 48% for the
0mm gap width compared to the 40% under-estimation at a 2mm grid size. The
dose predicted by Pinnacle below the closed leaf tips is strongly dependent on the
parameters used in the MLC model. As presented in a previous publication (Williams
and Metcalfe), the leaf tip radius used for our Pinnacle MLC model was 12cm instead
of the physical radius of 8cm. By decreasing the leaf tip radius of the Pinnacle MLC
model a better agreement between the measured and predicted end leaf leakage could
have been achieved but this would compromise the accuracy of the model in the
penumbral region of open MLC elds.
In Figure 9.5(b) the FWHM of the predicted and measured dose proles for the
gap widths are plotted. Based on the EBT lm measurements the FWHM for the
0.6mm gap width was 3.5mm at a depth of 1.5cm in solid water, with a maximum
transmission of 0.51cGy/MU. The amount of radiation transmitted through closed
MLC leaf pairs is not trivial, and for a single eld it could represent almost 25% of
the dose from the open portion of the eld being unintentionally delivered to a 3.5mm
wide strip of the patient. Pinnacle predicts the presence of these strips, however in
our case it under-estimated the magnitude of the end leaf leakage and subsequently
over-estimated the FWHM.
9.3.1.2 O-axis end leaf leakage
The o-axis dependence of the end leaf leakage was also investigated, and is illustrated
in Figure 9.6. end leaf leakage was measured at 5cm, 10cm and 15cm o axis in the
direction of leaf travel. Note that placement of closed leaf pairs 15cm o-axis is unlikely
in clinical situations. For the 0mm gap width the amount of radiation transmitted
through the closed leaf tips decreased as the gap was positioned further o-axis, for all
other leaf gaps the transmission was independent of the o-axis position. The shape
9.3. Results and discussion
190
of the transmitted prole for the 0mm gap at the furthest o-axis position can be
explained by the leaf geometry. The curvature of the MLC leaf is such that the top
and bottom points of the leading edge of the leaf are 4.54mm further back from the
leaf tip, which is 513.15mm from the source and has a total leaf thickness of 61.30mm.
Applying these parameters to the geometry shown in Figure 9.7 it was determined that
when the 0mm gap is positioned at distances greater than 148.2mm from the central
axis it is possible for ray-lines from the source to pass through both MLC leaf tips. The
attenuation by both leaves resulted in a dip in the transmitted radiation at the leaf
gap centred at 15.0cm o-axis. For the 0.6mm gap the corresponding position after
which attenuation by both leaves could have occurred was calculated to be 158.0mm.
The amount of transmitted radiation predicted by Pinnacle decreased as the oaxis distance increased. This was due in part to the use of the larger radius of curvature
for the leaf tip in the model. The leaf tip radius is used by Pinnacle to determine the
amount of attenuating material seen by each ray-line as it passes through the leaf.
The larger radius results in a greater thickness of leaf material being calculated and
hence the attenuation by the leaf tip is overestimated. The eects of the larger radius
are exaggerated at o-axis distances due to the oblique incidences of ray-lines on the
leaf tip producing longer path lengths through the leaves. We are unable to explain
why the Pinnacle predicted dose increased at an o axis distance of 15cm for the 0mm
gap width; however when the leaf end radius was change from 12cm to 8cm a uniform
peak dose was predicted at all o axis locations.
9.3.1.3 Clinical IMRT end leaf leakage
The measured and predicted planar dose distributions for the IMRT eld are shown
in Figure 9.8. The presence of closed leaf pairs was evident throughout the entire
eld, with end leaf leakage contributing dose at numerous locations across the eld.
Two specic locations were chosen for analysis; their positions have been indicated in
9.3. Results and discussion
191
Figure 9.6: O-axis end leaf leakage for the a) 0mm gap width b) 0.6mm gap width
and c) 3mm gap width
9.3. Results and discussion
192
Figure 9.7: The geometry of the Millennium MLC leaf was used to determine the
o-axis distances at which ray-lines from the source would begin to pass through both
leaf tips for the 0mm and 0.6mm leaf gaps.
9.3. Results and discussion
193
Figure 9.8: A wide IMRT eld (a) Radiographic EDR2 lm grey scale map at 10cm
depth in solid water (b) RTPS planar dose maps taken at 10cm depth in solid water
of a wide IMRT eld showing end leaf leakage. The lines shown represent where line
proles were taken.
Figure 9.8. Line 1 occurred inside a low intensity region overlying a critical structure,
and line 2 was in a higher intensity region of the eld that contributed dose to the
target volume. Figure 9.9(a) and (b) show the measured and predicted dose proles
across the end leaf leakage at positions 1 and 2 respectively. The doses in Figure 9.9
are the total doses resulting from all 25 fractions.
The maximum dose measured along line 1 was 3.04Gy and the surrounding low
intensity region received a dose of approximately 1.25Gy. The end leaf leakage increased the dose by 1.8Gy in the low intensity region. Pinnacle predicted the presence
and location of the leakage but under-estimated the amount. The maximum dose predicted by Pinnacle across the end leaf leakage was 2.4Gy; this was 20% lower than the
measured value. Across line 2 the maximum measured dose was 6.7Gy compared to
5.4Gy predicted by Pinnacle, again a 20% dierence. The dose in the uniform intensity
region surrounding the end leaf leakage at line 2 was approximately 3.95Gy; hence the
end leaf leakage contributed an additional 2.75Gy to the dose in this region.
9.3. Results and discussion
194
Figure 9.9: Proles taken across (a) Line 1 in a low intensity shielded region of the
IMRT eld shown in gure 8 and (b) Line 2 in a high intensity region of the eld.
9.3. Results and discussion
195
The dose that results from end leaf leakage can be substantial, with increases of
up to 1.8 and 2.75Gy recorded for this particular eld. These doses were measured at
a depth of 10cm in solid water for a single IMRT eld; the entire treatment consisted
of 7 beams delivering 50Gy to the target volume. Additional occurrences of end leaf
leakage were present in the IMRT eld and also in some of the other 6 beams. Doses of
the order of 2-3Gy observed here increase the risk of complication for healthy tissues
and critical structures. This is especially pertinent for IMRT which to achieve the
treatment objective relies heavily on the validity of tolerance doses specied for critical
structures and on the accuracy of the planning system in predicting the dose delivered
to those structures. For a serial structure such as the spinal cord an additional 2-3Gy
may increase the dose beyond an acceptable limit. The ability of a planning system
to predict this dose, even with limited accuracy, enables the location of the leakage to
be identied and the dose contribution to critical structures assessed and potentially
compensated for in the IMRT optimisation.
The data presented in this study is specic to the leaf model implemented at
Illawarra Cancer Care Centre, Wollongong Hospital and the Centre for Medical Radiation Physics, University of Wollongong. The magnitude of dose resulting from end
leaf leakage in a wide eld IMRT delivery using the Millennium MLC will be dependent
on many factors. The accuracy of MLC calibration and the minimum gap width specied will directly impact on the leakage; in our case we were using a 0.6mm minimum
leaf gap. The accuracy of a planning system in predicting the magnitude and distribution of dose from end leaf leakage is dependent on the beam model, MLC model, and
calculation grid. In its' present implementation of wide eld IMRT Pinnacle does not
selectively position closed leaf pairs, instead their position is governed by the leaf positions of neighboring open segments. As the IMRT eld widens the possible position
of the closed leaf pairs narrows and is more central in the eld. Possible improvements
9.4. Conclusion
196
would be to increment the position of closed leaf pairs across the eld, as suggested
by MacKenzie et al. (2002); or alternatively position the leaves in locations that do
not coincide with high risk critical structures.
9.4 Conclusion
The leakage of radiation through closed leaf ends of a Millennium MLC can contribute
a signicant amount of dose when positioned inside the eld. The maximum leakage
measured for a single eld was 0.39cGy/MU for a 0mm gap and 0.51cGy/MU for
a 0.6mm gap. In wide eld IMRT the closed leaf ends are distributed throughout
the central 29cm of the eld due to the 14.5cm over-travel limits of the MLC. Their
dosimetric contribution and anatomical location should not be ignored. For a single
IMRT eld end leaf leakage contributed an additional 2-3Gy over the course of treatment. The RTPS investigated in this study provided an elegant leaf model however
there were dierences in predicted versus measured end leaf leakage doses for clinical
situations of 20-40%.
In practice any signicant leakage predicted by the RTPS should be veried prior
to delivery of the treatment. If a sensitive serial structure, such as the spinal cord, is
being treated close to tolerance any extra dose unaccounted for in the plan may result
in unacceptable tissue complications. The ability to plan and treat IMRT elds wider
than 14.5cm with the Millennium MLC has improved the eciency and exibility of
IMRT treatments; however care needs to be taken to ensure that these gains do not
inadvertently compromise treatment ecacy.
Chapter 10
Summary and future work
10.1 Evaluation of advantages or disadvantages of
IMRT over 3DCRT for prostate radiotherapy
In Chapter 2, IMRT plans were compared with 3DCRT plans for 16 prostate patients.
IMRT resulted in lower rectal doses leading to reductions in rectal NTCPs for all
sixteen patients. The rectal dose reductions were seen over the whole dose range but
the magnitude of the reductions varied from patient to patient. The delivery eciency
was compared and this resulted in the IMRT plans requiring on average 42% more
MU for plan delivery. While dose distribution improvements are more important than
time eciency for cancer patients, quantifying the extra delivery time is necessary for
planning patient throughput.
197
10.2. Evaluation of biological optimisation tools for prostate IMRT
198
10.2 Evaluation of biological optimisation tools for
prostate IMRT
In Chapter 3, IMRT plans were created for sixteen prostate patients using biological
optimisation objectives. The maximum gEUD IMRT objective was used for the rectum. The gEUD function requires the desired maximum gEUD value and the value of
the parameter a. For each patient, three IMRT plans were created using a goal maximum gEUD value of as low as possible without compromising target coverage and a
values of 3, 4.5 and 9 for the rectum. While equivalent PTV dose was retained, the
rectal DVH changed depending on the value of a used. As the value of a increased, the
volumes receiving high doses were reduced and the volumes receiving mid-low doses
increased. Only the plans optimised with a = 3 resulted in statistically signicant
reductions in gEUD (calculated with a=3) over plans optimised with a=3 or a=9.
This suggests that the most gains in rectal dose reductions are in reducing volumes
receiving mid-low doses which is consistent with the anatomy of the rectal volume.
10.3 Investigation of Volumetric Modulated Arc
Radiotherapy for prostate cancer
Volumetric Modulated Arc Radiotherapy (VMAT) plans were investigated in Chapter
4. VMAT is a relatively new IMRT delivery technique whereby a modulated dose
distribution is delivered with a conventional linac using a continuously rotating gantry
with dynamic MLC motion. Recent literature suggests VMAT requires signicantly
fewer MU than static gantry angle IMRT for equivalent target coverage and normal
tissue sparing. For ten prostate patients, seven eld static gantry angle IMRT plans
were compared with VMAT plans planed with the same biological and physical dose
10.4. Optimisation of prostate IMRT plans based on the theoretical 'ideal dose'
199
objectives. For all ten patients, VMAT resulted in reductions in the volumes receiving
at 25Gy. This result was statistically signicant. VMAT required on average 18.6%
fewer MU than static gantry angle IMRT for delivery. In addition, the VMAT plans
were all limited to a 2 minute delivery time. This was in comparison to an average of
7.5minutes required for seven eld IMRT at this institution for rst beam on to last
beam o.
10.4 Optimisation of prostate IMRT plans based
on the theoretical 'ideal dose'
A method of reducing optimisation time by using prior knowledge of the optimal
dose distribution for prostate cancer IMRT was presented. The method generated
an optimal deliverable photon dose distribution based on the anatomy of the patient.
An 'ideal' DVH was then calculated for this optimal dose distribution from which
the gEUD was calculated. The gEUD value was used as the optimisation goal for a
maximum gEUD IMRT objective function. The optimisation algorithm was able to
return an equal or superior dose distribution in one half to one third the time required
compared that achieved by optimising without prior dose distribution knowledge.
10.5 Investigation of the dosimetric eect of rectal
balloon cavities
In Chapter 6, the eect of an air cavity on surrounding dose distributions was investigated for a commercial rectal balloon using radiochromic lm. The dosimetry was
performed under a number of irradiation conditions - single eld, 3DCRT, IMRT and
helical tomotherapy. For these four conditions the accuracy of two commercial RTPSs
10.6. Evaluation of in vivo dosimetry of the rectal wall using rectal balloons
combined with a novel MOSFET dosimeter
200
in calculating the dose surrounding rectal balloon cavities was investigated. The rectal
balloon cavity was found to perturb the dose in the same way as see in other air cavity
situations. That is, for beams incident on the cavity, a decrease in the anterior (in
direction of beam source) cavity wall dose due to loss of electronic equilibrium was
observed. A secondary dose build up at the posterior cavity wall was also observed.
For a beam incident laterally on the cavity, the posterior rectal wall dose increased due
to penumbral aring. These eects were calculated by the Pinnacle RTPS but underestimated. For seven eld 3DCRT and IMRT plans, the anterior rectal wall dose was
over-predicted and the posterior rectal wall dose under-predicted. This resulted in an
over-prediction by the Pinnacle RTPS of the rectal wall volumes receiving mid-high
doses. For the helical tomotherapy plan, the rectal DVH was accurately calculated.
In general, the secondary dose build up in the tissue beyond the balloon (in the
prostate target) was less than anticipated. It was reasonably accurately calculated by
both planning systems and increased condence that the air cavity produced by the
balloon does not produce loss of dose to the target. The fact that the Pinnacle RTPS
over-estimated the rectal DVH and the TomoTherapy RTPS slightly under-predicted
the rectal DVH suggests that rectal DVH cutpoints derived from conventional IMRT
cannot be directly transferred to the tomotherapy situation.
10.6 Evaluation of in vivo dosimetry of the rectal wall using rectal balloons combined with
a novel MOSFET dosimeter
In Chapter 7, the rectal balloon combined with a MOSFET for in vivo dosimetry was
investigated. The rectal balloon provides an excellent means for in vivo dosimetry of
the rectal wall in prostate radiotherapy. This provides both rectal wall dose measure-
10.6. Evaluation of in vivo dosimetry of the rectal wall using rectal balloons
combined with a novel MOSFET dosimeter
201
ments and possibly PTV dose measurements, as the anterior rectal wall is generally
contained by the PTV. The MOSkin detector was used. This is a MOSFET detector
with a reproducible build up of 70m WED. A number of MOSkin detectors were
placed on the outside of a commercial rectal balloon which was placed in a specically
designed phantom. A helical tomotherapy plan delivering 70Gy in 2.5Gy fractions to
a hypothetical prostate target was delivered.
The MOSkin detectors provided real time read out of dose with read out frequency
of 1Hz. The MOSkin measurements were compared with EBT lm measurements
and the TomoTherapy RTPS calculation. The MOSkin measured dose was less than
that of the RTPS at anterior locations. The dose to the anterior cavity wall was also
measured with Gafchromic EBT lm. The MOSkin under-responded compared with
the EBT lm which under-responded compared to the RTPS calculation. The EBT
lm measurement suggests an over-prediction of the anterior rectal wall dose by the
RTPS, in agreement with the results presented in Chapter 6. The MOSkin measured
dose was less than that of the EBT lm due to angular response, whereby a lower
sensitivity is observed for radiation incident through the Si substrate on the underside of the MOSkin. Two correction strategies were discussed to account for this. The
rst is a lter placed on the top of the MOSkin to reduce the sensitivity to radiation
incident from the top of the MOSkin. The second is a dual MOSkin conguration
whereby two MOSkins are placed face to face and the average reading of the two
detectors is taken. The dual MOSkin conguration proved to be a successful method
of reducing angular response to within 2.5$. The dual MOSkin was used to verify
the anterior rectal wall dose for a 3DCRT and IMRT plan for a hypothetical prostate
treatment in a phantom. The dual MOSkin-measured anterior rectal wall dose was
2.62% (3DCRT) and 3.17% (IMRT) lower than the Pinnacle RTPS calculated dose to
the dual MOSkin.
10.7. Evaluation of the MOSkin and Gafchromic EBT Film for clinical surface dose
verication
202
10.7 Evaluation of the MOSkin and Gafchromic
EBT Film for clinical surface dose verication
In Chapter 8, two novel skin dosimeters were investigated for clinical surface/skin
dose verication. Recent literature suggests that the TomoTherapy Hi-ART RTPS
over-predicts the surface dose. This is potentially a problem for cases where the surface/skin is part of the target volume, such as in total scalp irradiation. A novel
surface dosimeter, the MOSkin was compared with Gafchromic EBT Film for surface
dosimetry. Both detectors provided excellent spatial resolution in the depth direction.
The MOSkin and EBT lm were then used to verify the dose delivered to a phantom,
simulating a total scalp irradiation treatment. EBT lm was placed in the transverse
slices of an anthropomorphic phantom and on the surface of the phantom. MOSkin
detectors were placed on the surface in locations corresponding to the EBT lm locations. The RTPS surface dose calculation was found to be within 2% of the transverse
lm measurements. The RTPS surface dose calculation was assisted in this case by the
multiple overlapping beamlets delivered tangentially to the surface of the phantom,
thus increasing the surface dose. The surface doses measured with the EBT lm and
MOSkin were not compared with the plan as they represent two dierent measurement geometries. It was found that the MOSkin measured dose was consistently less
than the EBT lm measured dose, due to the shallower WED of the MOSkin. This
study demonstrated that the location of the sensitive volume, hence the WED of the
detector determines the measured surface dose.
10.8. Measurement of collimator leakage for a linac MLC
203
10.8 Measurement of collimator leakage for a linac
MLC
In Chapter 9, the use of a commercial MLC for wide eld MLC and implications of
end leaf leakage is discussed. One vendor's MLC has a maximum leaf over travel of
14.5cm. This means that for jaw dened elds larger than this, any opposite leaf pairs
cannot be joined with jaw shielding. A recent IMRT optimisation algorithm in the
Pinnacle RTPS allows for IMRT elds to be larger than 14.5cm therefore for elds
wider than this, opposite leaf pairs are joined without jaw shielding. This is a problem
as the leaf ends are rounded, therefore when they are joined, leakage occurs, termed
'end leaf leakage'. This leakage was measured for simple square elds and varying
distances between opposing leaves using radiographic and radiochromic lm. The lm
measurements were compared with the Pinnacle RTPS calculation of the leakage. It
was found that for opposing leaf gaps of < 5mm, the Pinnacle RTPS under-estimated
the magnitude and the FWHM of the leakage. The end leaf leakage was measured
for a clinical IMRT eld. An under-prediction by the Pinnacle RTPS of up to 2.75Gy
over a 50Gy treatment course was found.
10.9. Future work
204
10.9 Future work
10.9.1 Following on from the current work
At the completion of this thesis it was clear there were a number of things that time/resource constraints have prevented from being investigated. Topics worthy of future
investigation include:
Extend the planning studies presented in Chapters 2-4 to include patients with
seminal vesicle involvement.
Investigation of VMAT plans for other treatment sites, including head and neck
and pelvic nodal irradiation
Comparison of VMAT plans with helical tomotherapy plans
Performing the IMRT and VMAT planning studies on patients with rectal balloons inserted
Pursuing a small clinical trial investigating the use of the rectal balloon/MOSkin
apparatus for daily in vivo dosimetry
10.9.2 Prostate radiotherapy
Chapters 2-5 describe planning and delivery techniques to reduce OAR doses for
prostate radiotherapy. IMRT was shown to decrease rectal dose, compared with
3DCRT. Biological optimisation was then shown to be a useful method for reducing rectal doses for prostate IMRT. A new delivery technique, VMAT, was then used
with biological optimisation objectives to result in even further rectal dose reductions
with the added benet of large eciency gains. An algorithm was then presented that
allows for patient specic IMRT objectives to be set.
10.9. Future work
205
10.9.3 Target denition
Chapters 2-4 present planning studies investigating the eects of dierent delivery
and optimisation techniques on prostate radiotherapy. In these planning studies, the
target was dened as the prostate only. However, a proportion of prostate radiotherapy
patients require irradiation of the seminal vesicles (SVs) in addition to the prostate.
Denition of the SVs in the target volume increase the complexity of the target shape.
It follows then that the advantages of complex delivery techniques such as IMRT,
VMAT and helical tomotherapy over 3DCRT could be increased when the SVs are
included in the target volume.
The advantages of IMRT over 3DCRT and the ecacy of VMAT when treating
the prostate plus SVs is currently being investigated. The SVs were contoured on two
of the patients used in Chapter 2. The SVs were contoured according to the PROFIT
(Ontario Clinical Oncology Group (OCOG) NCT00304759) clinical trial guidelines; the
SVs are contoured from the base of the prostate to 1cm superior to the prostate base.
IMRT and 3DCRT plans were then created based on the same objectives presented in
Chapter 2.
Figures 10.1-10.4 show the dose distributions and cumulative DVHs for the prostate
plus SVs, rectum, bladder and femoral heads. It is clear from Figures 10.2 and 10.4
that IMRT results in signicantly reduced rectal and bladder doses. The rectal dose
reductions are increased compared with that seen with the same patients treating to
the prostate only (Chapter 2). It is expected that the rectal and bladder dose cut points
used clinically at Illawarra Cancer Care Centre may not be met with the 3DCRT plans
when treated to 78Gy. Therefore a major benet of IMRT for these patients could be
the ability to dose escalate to 78Gy whilst meeting rectal and bladder constraints.
The remainder of the 16 patients used in Chapter 2 will be re-planned to deliver
78Gy to the prostate and SVs using 3DCRT and IMRT techniques.
10.9. Future work
206
Figure 10.1: Dose distributions for (a) 3DCRT (sagittal) (b) IMRT (sagittal) (c)
3DCRT (transverse) and (d) IMRT (transverse) plans for Patient 5 including seminal
vesicles
10.9. Future work
207
Figure 10.2: Cumulative DVHs for (a) PTV and rectum and (b) bladder and femoral
heads for Patient 5
10.9. Future work
208
Figure 10.3: Dose distributions for (a) 3DCRT (sagittal) (b) IMRT (sagittal) (c)
3DCRT (transverse) and (d) IMRT (transverse) plans for Patient 6 including seminal
vesicles
10.9. Future work
209
Figure 10.4: Cumulative DVHs for (a) PTV and rectum and (b) bladder and femoral
heads for Patient 6
10.9. Future work
210
10.9.3.1 Target motion
The planning techniques evaluated in this thesis were performed on planning CT scans
taken prior to the start of treatment. There have been many reports of large rectal
volume dierences and prostate movement between fractions (Bylund et al. , 2008;
Hsi et al. , 2008; Huang et al. , 2002a; Osei et al. , 2009). As a result, the plan
obtained on the planning system prior to treatment is generally not what is obtained
during the treatment. Various methods are used to reduce the interfractional and
intrafractional anatomical variations, such as attempts to control bowel and bladder
lling and the use of rectal balloons McGary et al. (2002); Teh et al. (2001, 2002);
Wachter et al. (2002). Methods also exist to take into account for prostate motion
such as daily volumetric imaging or use of ducial markers to track the prostate.
These methods allow targeting of the prostate immediately prior to the delivery of
each fraction. However, a recent report by Noel et al. (2009) found that neither preor post-irradiation imaging were sucient to predict intrafractional prostate motion.
This suggests that prostate immobilisation or real-time prostate tracking should be
employed.
The dierences in anatomy from the planning CT during the course of the treatment results in a loss of condence that the planned dose distribution is actually
delivered. Two approaches to minimising this problem should be considered. The rst
is to immobilise the prostate for the planning CT and subsequent fraction delivery.
This is achieved using a rectal balloon. Currently, there are few, if any institutions
in Australia that employ rectal balloons, even though they have been shown to immobilise the prostate and reduce rectal toxicity (D'Amico et al. , 2006; van Lin et al.
, 2005b, 2007; McGary et al. , 2002; Patel et al. , 2003; Teh et al. , 2005; Wang
et al. , 2007). The employment of rectal balloons would increase condence that the
anatomy for each fraction delivery is the same as the anatomy seen on the planning
10.9. Future work
211
CT. Disadvantages of this approach are the relatively high cost of rectal balloons (relative to treatment cost in Australia) and the discomfort for the patient. There have
been reports however, that show generally good patient acceptance of rectal balloons
(Goldner et al. , 2006).
The second approach is to take daily volumetric images of the target region and
fuse these to the planning CT data. If large enough variations in the anatomy are
observed, then the registration data used for fusion of the image sets could be used
to modify the uence maps of each IMRT eld, hence the individual segments could
be modied to account for the changed anatomy (Court et al. , 2005; Ludlum et al. ,
2007). This method, termed 'adaptive radiotherapy' would work as follows (Yan et al.
, 1997, 1998):
Planning CT taken, anatomy delineated and IMRT plan created
For each fraction, patient has volumetric image which is fused to planning CT
Dosimetric error due to anatomical variations between planned and daily CT are
calculated
If the dosimetric error is outside of set tolerance values, the IMRT segments are
modied to achieve desired dose distribution
New IMRT plan is checked by oncologist and delivered
This approach would require a modied optimisation algorithm, one that would
take the original, planned solution and work only to modify it. A fast dose calculation
engine would also be required, as the patient would be on the couch waiting for delivery.
A more recent report on direct aperture optimisation for adaptive radiotherapy by
Mestrovic et al. (2007) showed that with their algorithm, the total treatment time
would only be increased by 2, 4 or 6s for prostate movements of 0.25, 0.5 and 0.75cm
212
10.9. Future work
respectively. This approach has the disadvantage that intrafractional motion is not
taken into account. Therefore ideally, it should be coupled with a rectal balloon, so that
intrafractional motion is minimised. Another disadvantage is the extra imaging dose
that the patient receives. This is already being delivered in institutions that employ
daily pre-treatment volumetric imaging yet the eects of this extra imaging dose are
not well known. This could be negated if volumetric imaging modalities that don't
use ionising radiation, such as Magnetic Resonance Imaging (MRI) or Ultrasound, are
employed for daily pre-treatment imaging. Already, prototype combined MRI/linacs
are being tested (Lagendijk et al. , 2008; Kirkby et al. , 2008).
The accuracy of this approach is also dependent on the accuracy of deformable
registration algorithms, which are required to track the dose to individual volume
elements in the patient. This is necessary so that daily delivered doses and DVHs
can be added and tracked for possible treatment modication based on the delivered
doses.
10.9.4
In vivo
dosimetry
This thesis has investigated the use of a novel MOSFET detector, the MOSkin. All
of the work contained in this thesis however is performed on a phantom. To realise
the full potential of the MOSkin, measurements on patients need to be performed.
The CMRP is involved in two projects to implement in vivo MOSkin use to monitor
patient dose:
Skin measurements: Skin toxicity during breast radiotherapy is a common and
uncomfortable side eect. The MOSkin would provide a useful measurement of the
skin dose to the breast, particularly if breast immobilisation devices are introduced.
As skin toxicity develops over the treatment fraction, the depth of the radiosensitive
basal layer in the skin increases. Therefore it is proposed that a series of MOSkin
10.9. Future work
213
detectors is provided, covering a range of WEDs that can be used depending on at
what depth the oncologist wants to measure the dose.
Rectal wall measurements: The dual MOSkin approach described in Chapter
7 is being actively pursued for implementation in a commercial rectal balloon. The
MOSkin detector would be placed, during manufacturing, in the anterior wall of the
balloon. This would provide a dual MOSkin measurement at the rectal wall for target
and rectal wall dose monitoring. It has been proposed by that the readout system
and supply bias voltage would be miniaturised and placed on the treatment couch,
connected to the MOSkin/Balloon apparatus. This would then be connected with a
wireless connection (possibly via Bluetooth) with a laptop computer outside of the
treatment bunker and the MOSkin read out during the treatment delivery. Software
on the laptop would control readout and storage of measured doses during the course of
the treatment. One possible problem would be calibration of the MOSkin detectors, as
they would be supplied already embedded in the rectal balloon, meaning the MOSkins
would have to be pre-calibrated prior to integration in the balloon. However, the
MOSkin technology developed at the CMRP has resulted in a reproducible sensitivity
between each detector in a single batch ( 2%). This allows a single sensitivity for a
whole batch of MOSkins (all from the same Si wafer).
One issues that need to be addressed for the MOSkin to be used for real-time in vivo
measurements on patients is the temperature dependence of the MOSkin detector. It
has been shown that the MOSkin threshold voltage and response varies with temperature. A method to correct for this without dual MOSFE approach (Soubra et al.
, 1994; Thomson, 1987) and additional temperature sensors placed on a MOSFET
carrier has been proposed and justied by A.B. Rosenfeld and realised in a prototype
MOSFET reader at the CMRP (Safavi-Naeni, 2005). This will be incorporated into
the next generation of CMRP-developed MOSFET read out systems.
10.9. Future work
214
10.9.5 Summary
This thesis represents a combined study that may, in small increments, contribute
towards achieving optimal treatment planning, delivery and dosimetry methods for
radiotherapy for cancer patients. The two priorities for continuing the work undertaken
in this thesis are
Investigate IMRT, biological IMRT optimisation and VMAT planning and delivery techniques for prostate patients with seminal vesicle involvement
Clinical trials using the MOSkins on radiotherapy patients for skin dosimetry
and rectal wall dosimetry
Appendices
215
Appendix A
Ideal dose script
Chapter 5 describes a method for calculating the optimal dose for an individual patient
anatomy. This so called 'ideal dose' can then be used as a guide for IMRT optimisation.
The optimal dose is generated using a script (described below), which takes as input
a CT scan with contours in the CERR format.
A.1 Ideal dose calculation script
1
function [ i d e a l D o s e u n i q u e S l i c e s regParamS ] = g e t I d e a l D o s e ( planC , zone100
, dose100 , zone95 , dose95 , zoneZero , doseZero , PTV)
2
3 % c r e a t i o n o f i n d i v i d u a l dose g r i d s f o r each contour :
4
5
6 % c r e a t e 100% zone :
7
8 % g e t t h e CT data s e t s i z e , s l i c e numbers c o n t a i n i n g t h e 100% contour , a
9
mask o f 1 s and 0 s r e p r e s e n t i n g t h e contour and a dose g r i d f o r 100%
contour :
[ c t S i z e u n i q u e S l i c e s 1 0 0 d a t a S e t newDose100 ] = getMask ( planC , zone100 ,
d os e 1 00 ) ;
10
11 % c o n v e r t t o i n t 1 6 t o c o n s e r v e memory :
12 newDose100 = i n t 1 6 ( newDose100 ) ;
13
14 % save 100% dose g r i d and remove from workspace t o c l e a r out o f memory :
15 save ( ' newDose100 . mat ' , ' newDose100 ' ) ;
216
A.1. Ideal dose calculation script
16
17
217
c l e a r newDose100
18
19
20 % c r e a t e 95% zone :
21 [ c t S i z e u n i q u e S l i c e s 9 5 d a t a S e t newDose95 ] = getMask ( planC , zone95 , d o s e 95
);
22 newDose95 = i n t 1 6 ( newDose95 ) ;
23 save ( ' newDose95 . mat ' , ' newDose95 ' ) ;
24 c l e a r newDose95
25
26
27
28 % c r e a t e 0% ( s c a t t e r ) zone :
29 [ c t S i z e u n i q u e S l i c e s Z e r o d a t a S e t newDoseZero ] = getMask ( planC , zoneZero ,
doseZero ) ;
30 newDoseZero = i n t 1 6 ( newDoseZero ) ;
31 save ( ' newDoseZero . mat ' , ' newDoseZero ' ) ;
32 c l e a r newDoseZero
33
34
35
36 % Get i n f o r m a t i o n on t h e l o w e s t and h i g h e s t s l i c e #s c o n t a i n i n g t h e
37 % c o n t o u r s . This i s needed l a t e r when re sampling t h e dose cube :
38 m i n S l i c e ( 1 , 1 ) = min ( u n i q u e S l i c e s 1 0 0 ) ;
39 m i n S l i c e ( 1 , 2 ) = min ( u n i q u e S l i c e s 9 5 ) ;
40 m i n S l i c e ( 1 , 3 ) = min ( u n i q u e S l i c e s Z e r o ) ;
41 l o w e s t S l i c e = min ( m i n S l i c e ) ;
42
43 m a x S l i c e ( 1 , 1 ) = max( u n i q u e S l i c e s 1 0 0 ) ;
44 m a x S l i c e ( 1 , 2 ) = max( u n i q u e S l i c e s 9 5 ) ;
45 m a x S l i c e ( 1 , 3 ) = max( u n i q u e S l i c e s Z e r o ) ;
46 h i g h e s t S l i c e = max( m a x S l i c e ) ;
47
48
49 % c r e a t e penumbral mask :
50
51 penThick = 1 ; % t h i c k n e s s in cm o f penumbral r e g i o n around PTV
52
53 % g e t mask o f 1 s t h r o u g h t o 10 s r e p r e s e n t i n g d i f f e r e n t dose l e v e l s in t h e
penumbra :
54 penMask = getPenMask ( planC , PTV, penThick , c t S i z e ) ;
55 % c o n v e r t t o i n t 8 t o save memory
56 penMask = i n t 8 ( penMask ) ;
57
A.1. Ideal dose calculation script
218
58 % remove c e n t r e o f penumbral zone ( which i s 100%/95% zone i . e . PTV) and
r e p l a c e w i t h 0Gy . The PTV r e g i o n in t h e penumbral mask w i l l have a
mask v a l u e o f 10 t h e r e f o r e t h i s v a l u e i s s u b t r a c t e d from t h e c e n t r e
o f t h e penumbral mask :
59
60 d o s e C e n t r e = 10;
% 10 i s t h e mask v a l u e o f t h e PTV
61
62 % g e t mask o f PTV and s e t dose t o e q u a l doseCentre . This r e s u l t s in a
63
mask o f 0 s and doseCentre v a l u e s
[ c t S i z e u n i q u e S l i c e s d a t a S e t newDoseCentreDose ] = getMask ( planC , PTV,
doseCentre ) ;
64
65 % c o n v e r t t o i n t 8 t o save memory
66 newDoseCentreDose = i n t 8 ( newDoseCentreDose ) ;
67
68 % add t h e o r i g i n a l penumbral mask t o t h e newDoseCentreDose t o r e s u l t in a
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
penumbral mask w i t h 1 s t h r o u g h t o 10 s b u t 0 i n s i d e and o u t s i d e o f
t h e penumbra
penMask = penMask+newDoseCentreDose ;
% c r e a t e t h e penumbral dose g r i d :
% penumbral r e g i o n d o s e s . Taken r o u g h l y from a 10x10cm f i e l d penumbra :
region1 = 1560;
region2 = 2465;
region3 = 3104;
region4 = 3861;
region5 = 4618;
region6 = 5327;
region7 = 5936;
region8 = 6435;
region9 = 6770;
region10 = 6942;
% i n i t i a l i s e penumbral dose g r i d :
penDose = zeros ( c t S i z e ( 1 , 1 ) , c t S i z e ( 1 , 2 ) , c t S i z e ( 1 , 3 ) ) ;
clear i
j k
% a s s i g n d o s e s t o penumbral r e g i o n s . Each v o x e l in t h e penumbral mask i s
i n t e r r o g a t e d and t h e a p p r o p r i a t e r e g i o n dose i s a p p l i e d :
f o r k=1: c t S i z e ( 1 , 3 )
f o r i =1: c t S i z e ( 1 , 1 )
f o r j =1: c t S i z e ( 1 , 2 )
i f penMask ( i , j , k ) == 1
penDose ( i , j , k ) = r e g i o n 1 ;
e l s e i f penMask ( i , j , k ) == 2
penDose ( i , j , k ) = r e g i o n 2 ;
e l s e i f penMask ( i , j , k ) == 3
penDose ( i , j , k ) = r e g i o n 3 ;
219
A.1. Ideal dose calculation script
101
102
103
104
105
106
107
108
109
110
111
112
113
114
e l s e i f penMask ( i , j , k ) == 4
penDose ( i , j , k ) = r e g i o n 4 ;
e l s e i f penMask ( i , j , k ) == 5
penDose ( i , j , k ) = r e g i o n 5 ;
e l s e i f penMask ( i , j , k ) == 6
penDose ( i , j , k ) = r e g i o n 6 ;
e l s e i f penMask ( i , j , k ) == 7
penDose ( i , j , k ) = r e g i o n 7 ;
e l s e i f penMask ( i , j , k ) == 8
penDose ( i , j , k ) = r e g i o n 8 ;
e l s e i f penMask ( i , j , k ) == 9
penDose ( i , j , k ) = r e g i o n 9 ;
e l s e i f penMask ( i , j , k ) == 10
115
end
116
end
117
end
118
end
119
end
120
end
121
end
122
end
123
end
124
end
125
end
126
end
127 end
128
129 save ( ' penDose . mat ' , ' penDose ' ) ;
130 % c o n v e r t t o i n t 1 6 t o save memory
131 penDose = i n t 1 6 ( penDose ) ;
132
133
penDose ( i , j , k ) =
region10 ;
134
135 % Combine t h e d o s e s :
136 load newDose100 . mat ;
137 load newDose95 . mat ;
138 load newDoseZero . mat ;
139
140 % g e t s i z e s o f each contour ' s dose g r i d :
141 s i z e 1 0 0 = s i z e ( newDose100 ) ;
142 s i z e 9 5 = s i z e ( newDose95 ) ;
143 s i z e z e r o = s i z e ( newDoseZero ) ;
144 s i z e p e n = s i z e ( penDose ) ;
145
146 % add each contour ' s dose g r i d t o g e t t o t a l dose g r i d :
147 i d e a l D o s e = newDose100 + newDose95 + newDoseZero + penDose ;
148
149 d o s e E r r o r = zeros ( c t S i z e ( 1 , 1 ) , c t S i z e ( 1 , 2 ) , c t S i z e ( 1 , 3 ) ) ;
150 d o s e E r r o r = i n t 1 6 ( d o s e E r r o r ) ;
220
A.1. Ideal dose calculation script
151
152
153
154
newDoseError = d o s e E r r o r ( : , : , l o w e s t S l i c e : h i g h e s t S l i c e ) ;
d o s e E r r o r = newDoseError ;
c l e a r newDoseError
155
156 % resample dose g r i d t o save memory :
157
158 % remove s l i c e s w i t h o u t dose v a l u e s :
159 newIdealDose = i d e a l D o s e ( : , : , l o w e s t S l i c e : h i g h e s t S l i c e ) ;
160
161 newSize = s i z e ( newIdealDose )
162
163 i d e a l D o s e = newIdealDose ;
164
165 c l e a r i j k newIdealDose
166
167 % now t o r e p l a c e ( or add ) dose d i s t r i b u t i o n in planC f i l e and save t h e
plan :
168
169 % t h i s data i s r e q u i r e d f o r CERR t o d i s p l a y and use t h e dose cube :
170 f r a c t i o n G r o u p I D = ' i d e a l ' ;
171 d o s e U n i t s = ' cGy ' ;
172 d o s e E d i t i o n = 1 ;
173 d e s c r i p t i o n = ' i d e a l d o s e d i s t r i b u t i o n ' ;
174 r e g i s t e r = ' non CT ' ;
175
176 regParamS . h o r i z o n t a l G r i d I n t e r v a l = planC f 1 , 3 g . u n i f o r m S c a n I n f o . g r i d 1 U n i t s ;
177
178
179
180
181
182
183
184
% ( x v o x e l w idt h )
regParamS . v e r t i c a l G r i d I n t e r v a l =
% ( y v o x e l w id th )
planC f 1 , 3 g . u n i f o r m S c a n I n f o . g r i d 2 U n i t s ;
regParamS . c o o r d 1 O F F i r s t P o i n t = planC f 1 , 3 g . u n i f o r m S c a n I n f o . x O f f s e t + planC
f 1 , 3 g . u n i f o r m S c a n I n f o . g r i d 1 U n i t s planC f 1 , 3 g . u n i f o r m S c a n I n f o .
g r i d 1 U n i t s /2
planC f 1 , 3 g . u n i f o r m S c a n I n f o . g r i d 1 U n i t s ( c t S i z e ( 1 , 1 ) / 2 )
% ( x v a l u e o f c e n t e r o f upper l e f t v o x e l on a l l s l i c e s )
regParamS . c o o r d 2 O F F i r s t P o i n t = ( planC f 1 , 3 g . u n i f o r m S c a n I n f o . y O f f s e t +
planC f 1 , 3 g . u n i f o r m S c a n I n f o . g r i d 2 U n i t s /2
( c t S i z e ( 1 , 2 ) / 2 ) planC f 1 , 3 g .
u n i f o r m S c a n I n f o . g r i d 2 U n i t s ) % ( y v a l u e o f c e n t e r o f upper l e f t v o x e l
on a l l s l i c e s )
clear i
j k
f o r i=l o w e s t S l i c e : h i g h e s t S l i c e
regParamS . z V a l u e s ( 1 , i +1 l o w e s t S l i c e ) = planC f 1 , 3 g . s c a n I n f o ( 1 , i ) .
zValue ; %( z v a l u e s % o f a l l s l i c e s )
185 end
186
187 % t h i s f u n c t i o n adds t h e dose cube t o t h e plan :
188 planC = addDoseToPlan nick ( planC , i d e a l D o s e , d o s e E r r o r , f r a c t i o n G r o u p I D ,
d o s e E d i t i o n , d e s c r i p t i o n , r e g i s t e r , regParamS ) ;
189
190 % Saving t h e plan f o r v i e w i n g in CERR:
A.1. Ideal dose calculation script
191
[ saveFileName , savePathName , s a v e F i l t e r I n d e x ] = u i p u t f i l e ( ' . mat ' , ' Save
plan as : ' ) ;
l o c a t i o n = [ savePathName saveFileName ] ;
save ( l o c a t i o n , ' planC ' )
192
193
194
195 end
221
Appendix B
Monte Carlo simulations
B.1 Overview of simulations
This appendix describes the parameters used for the Monte Carlo simulations contained in this thesis. All simulations performed by the author were done using the
EGSnrc/BEAMnrc package Rogers et al. (1995). The versions used were BEAMnrcMP 2006 and 2007. The BEAMnrc package is an 'add-on' to the EGSnrc package to
allow for simulation of medical linear accelerators. That is, it contains input sources
and geometry designed specically for simulating medical linacs. Simulation begins
with the construction of the linac head with a range of component modules (CMs)
that describe the geometry of specic linac head components. The radiation transport
is controlled using the EGSnrcMP code.
The process of simulation was:
Creation of photon beam
Transport of electrons and photons through the linac head geometry, including
secondary collimators (jaws and MLC)
Collection of phase space data after linac head
222
223
B.1. Overview of simulations
Transport of contents of phase space le through phantom geometry (if dose
calculation was required)
The specic details of the simulation is now described.
Creation of photon beam: The photon beam was created by simulating
bremsstrahlung in a target. The initial electron beam characteristics incident on
the target were determined using trial and error to obtain a photon spectrum that
matched that of a Varian 21EX linac, used in all experiments. The photon spectrum
was matched by comparing the simulated depth dose curve with the (ion chamber)
measured depth dose curve for a 10x10cm photon eld. The nal electron beam used
was a circular beam with a 2-D Gaussian x-y distribution with a geometrical FWHM
of 0.13cm. The electron beam had a Gaussian energy spread given in Figure B.1 that
was peaked at 6.2MeV with an energy FWHM of 3%. The number of electrons used
was generally 50 x 10 .
Directional bremsstrahlung splitting (DBS) was used as a variance reduction technique, with a splitting number N=1000 Kawrakow et al. (2004). VR techniques are
used to increase the eciency of the simulation. DBS was used to increase the number
of bremsstrahlung photons created in the target. At each bremsstrahlung interaction
site, N bremsstrahlung photons are created, each with a statistical weight of 1/N.
Transport through the linac head geometry: The following components of
the linac head were constructed using the CMs given in brackets: Target (SLABS),
Primary Collimator (CONS3R), Vacuum Window (SLABS), Flattening Filter (FLATFILT), Monitor Chamber (CHAMBER), Mirror (MIRROR), Jaws (JAWS) and Multileaf Collimator (DYNVMLC) (Heath & Seuntjens, 2003). The geometry is given at
the end of this appendix in the example input le.
BEAMnrcMP allows energy thresholds to be set. When a particles energy decreases below the threshold, the particle is terminated and all its remaining energy
2
6
B.1. Overview of simulations
224
Figure B.1: Incident electron energy spectrum
is deposited at its current location. For surface dose simulations (Chapter 8), the
thresholds ECUT (for electrons) and PCUT (for electrons) were set to 0.521MeV and
0.01MeV respectively. For all other simulations, ECUT and PCUT were 0.7MeV and
0.01MeV respectively.
An example BEAMnrcMP input le is given in Section B.2.
Collection of Phase Space le: A phase space le was collected at the distal
end of the air gap between the linac head and the phantom. The phase space le is a
le that contains the energy, position, momentum vector and charge of every particle
crossing a designated plane. The phase space le can then be analysed using the
BEAMDP utility to obtain energy, uence and angular distrubution information.
Transport of the contents of the phase space le through phantom geometry:
The contents of the phase space le is used as the input for a second
B.1. Overview of simulations
225
simulation using DOSXYZnrc. DOSXYZnrc is a package that allows for dose scoring
in cartesian geometry. The geometry can either be dened manually by the user or
created from a CT data set. To create a phantom, the size, material and scoring voxel
resolution must be dened. The incident radiation can either be dened by the user
as either photon or electron beam or can be obtained using the contents of a phase
space le. An example DOSXYZnrc input le is given in Section B.3.
B.2. Example BEAMnrc input le
226
B.2 Example BEAMnrc input le
Varian 21EX Linear Accelerator #!GUI1.0
AIR521ICRU
0, 0, 0, 0, 0, 3, 1, IWATCH ETC. (OUTPUT OPTIONS)
50000000, 225, 63, 999, 2, 1000, 2, 0, # OF HISTORIES, RANDOM NUMBER SEEDS ETC.
20, 100, 4, 20, 1, 11.25, DIRECTIONAL BREM SPLITTING OPTIONS
-1, 19, -0.13, 0, 0, 1, 0.0, 0.0, 0.0, 0.0, DESCRIPTION OF SOURCE
1, SPECTRUM
/home/nick/HEN HOUSE/spectra/test.spectrum LOCATION OF ELECTRON SPECTRUM
0
0, 0, 0.521, 0.01, 0, 0, , 0 , ECUT,PCUT,IREJCT,ESAVE (ENERGY CUTOFF INFORMATION)
0, 0, 0, 0, 0, PHOTON FORCING (ANOTHER VARIANCE REDUCTION TECHNIQUE)
1, 9, SCORING INPUT
1, 1
30, 0, DOSE COMPONENTS
0.0, Z TO FRONT FACE
*********** start of CM SLABS with identifier Target ***********
2, RMAX
Tungsten Target
2, NSLABS
0, ZMIN
0.0889, 0.521, 0.01, 1, 1, 0
W521ICRU
0.1575, 0.521, 0.01, 2, 2, 0
CU521ICRU
*********** start of CM CONS3R with identifier PRIMCOLL ***********
20, RMAX
Primary Collimator
1.6, ZMIN
6, ZTHICK
2, NUM NODE
1.6, 0.4,
7.6, 1.9,
0.521, 0.01, 1, 23, 0,
AIR521ICRU
0.521, 0.01, 3, 3, 0,
STEEL521ICRU
*********** start of CM SLABS with identifier VACWIN ***********
20, RMAX
Vacuum Window
2, NSLABS
7.6, ZMIN
1.4, 0.521, 0.01, 0, 23, 0
AIR521ICRU
0.025, 0.521, 0.01, 1, 4, 0
BE521ICRU
*********** start of CM FLATFILT with identifier FLATFILT ***********
3.81, RMAX
Flattening Filter
9.0254, ZMIN
B.2. Example BEAMnrc input le
19, NUMBER OF LAYERS
1, 0.028, # CONES, ZTHICK OF LAYER 1
0,
0.064,
1, 0.028, # CONES, ZTHICK OF LAYER 2
0.064,
0.127,
1, 0.038, # CONES, ZTHICK OF LAYER 3
0.127,
0.191,
1, 0.041, # CONES, ZTHICK OF LAYER 4
0.191,
0.254,
1, 0.074, # CONES, ZTHICK OF LAYER 5
0.254,
0.381,
1, 0.135, # CONES, ZTHICK OF LAYER 6
0.381,
0.508,
1, 0.109, # CONES, ZTHICK OF LAYER 7
0.508,
0.635,
1, 0.109, # CONES, ZTHICK OF LAYER 8
0.635,
0.762,
1, 0.112, # CONES, ZTHICK OF LAYER 9
0.762,
0.889,
1, 0.104, # CONES, ZTHICK OF LAYER 10
0.889,
1.016,
1, 0.208, # CONES, ZTHICK OF LAYER 11
1.016,
1.27,
1, 0.191, # CONES, ZTHICK OF LAYER 12
1.27,
1.524,
1, 0.185, # CONES, ZTHICK OF LAYER 13
1.524,
1.778,
1, 0.168, # CONES, ZTHICK OF LAYER 14
1.778,
2.032,
1, 0.155, # CONES, ZTHICK OF LAYER 15
2.032,
2.286,
1, 0.142, # CONES, ZTHICK OF LAYER 16
2.286,
2.54,
1, 0.13, # CONES, ZTHICK OF LAYER 17
2.54,
2.794,
227
B.2. Example BEAMnrc input le
3, 0.097, # CONES, ZTHICK OF LAYER 18
2.794, 3.365, 3.81,
3.061, 3.302, 3.81,
1, 0.152, # CONES, ZTHICK OF LAYER 19
3.81,
3.81,
0.521, 0.01, 5, 5,
Copper
0.521, 0.01, 1, 23,
AIR521ICRU
0.521, 0.01, 5, 5,
Copper
0.521, 0.01, 1, 23,
AIR521ICRU
0.521, 0.01, 5, 5,
Copper
0.521, 0.01, 1, 23,
AIR521ICRU
0.521, 0.01, 5, 5,
Copper
0.521, 0.01, 1, 23,
AIR521ICRU
0.521, 0.01, 5, 5,
Copper
0.521, 0.01, 1, 23,
AIR521ICRU
0.521, 0.01, 5, 5,
Copper
0.521, 0.01, 1, 23,
AIR521ICRU
0.521, 0.01, 5, 5,
Copper
0.521, 0.01, 1, 23,
AIR521ICRU
0.521, 0.01, 5, 5,
Copper
0.521, 0.01, 1, 23,
AIR521ICRU
0.521, 0.01, 5, 5,
Copper
0.521, 0.01, 1, 23,
AIR521ICRU
0.521, 0.01, 5, 5,
Copper
0.521, 0.01, 1, 23,
AIR521ICRU
0.521, 0.01, 5, 5,
Copper
0.521, 0.01, 1, 23,
AIR521ICRU
0.521, 0.01, 5, 5,
Copper
228
B.2. Example BEAMnrc input le
0.521, 0.01, 1, 23,
AIR521ICRU
0.521, 0.01, 5, 5,
Copper
0.521, 0.01, 1, 23,
AIR521ICRU
0.521, 0.01, 5, 5,
Copper
0.521, 0.01, 1, 23,
AIR521ICRU
0.521, 0.01, 5, 5,
Copper
0.521, 0.01, 1, 23,
AIR521ICRU
0.521, 0.01, 5, 5,
Copper
0.521, 0.01, 1, 23,
AIR521ICRU
0.521, 0.01, 5, 5,
Copper
0.521, 0.01, 1, 23,
AIR521ICRU
0.521, 0.01, 5, 5,
Copper
0.521, 0.01, 1, 23,
AIR521ICRU
0.521, 0.01, 5, 5,
Copper
0.521, 0.01, 1, 23,
AIR521ICRU
0.521, 0.01, 5, 5,
Copper
0.521, 0.01, 1, 23,
AIR521ICRU
*********** start of CM CHAMBER with identifier CHAMBER ***********
10, RMAX
Monitor Chamber, 3 Windows, 4 signal plates
12.81728, ZMIN
0, 12, 0, NTOP, NCHM, NBOT
9.9, 9.95, 10, RADII FOR CENTRAL PART
0.0127, 0, ZTHICK, FLAG FOR LAYER 1 IN CENTRAL PART
0.521, 0.01, 6, 6,
KAPTON521ICRU
0.238, 0, ZTHICK, FLAG FOR LAYER 2 IN CENTRAL PART
0.521, 0.01, 7, 6,
AIR521ICRU
0.00508, 0, ZTHICK, FLAG FOR LAYER 3 IN CENTRAL PART
0.521, 0.01, 6, 6,
KAPTON521ICRU
0.239, 0, ZTHICK, FLAG FOR LAYER 4 IN CENTRAL PART
0.521, 0.01, 7, 6,
AIR521ICRU
229
B.2. Example BEAMnrc input le
0.00508, 0, ZTHICK, FLAG FOR LAYER 5 IN CENTRAL PART
0.521, 0.01, 6, 6,
KAPTON521ICRU
0.238, 0, ZTHICK, FLAG FOR LAYER 6 IN CENTRAL PART
0.521, 0.01, 7, 6,
AIR521ICRU
0.0127, 0, ZTHICK, FLAG FOR LAYER 7 IN CENTRAL PART
0.521, 0.01, 6, 6,
KAPTON521ICRU
0.238, 0, ZTHICK, FLAG FOR LAYER 8 IN CENTRAL PART
0.521, 0.01, 7, 6,
AIR521ICRU
0.00508, 0, ZTHICK, FLAG FOR LAYER 9 IN CENTRAL PART
0.521, 0.01, 6, 6,
KAPTON521ICRU
0.477, 0, ZTHICK, FLAG FOR LAYER 10 IN CENTRAL PART
0.521, 0.01, 7, 6,
AIR521ICRU
0.0127, 0, ZTHICK, FLAG FOR LAYER 11 IN CENTRAL PART
0.521, 0.01, 6, 6,
KAPTON521ICRU
0.635, 0, ZTHICK, FLAG FOR LAYER 12 IN CENTRAL PART
0.521, 0.01, 7, 6,
AIR521ICRU
0.521, 0.01, 6, 6, chamber wall
AIR521ICRU
0.521, 0.01, 6, 6, gap
AIR521ICRU
0.521, 0.01, 6, 6, container
AIR521ICRU
0, MRNGE
*********** start of CM MIRROR with identifier MIRROR ***********
7, RMAX
Mylar mirror at angle 35 degrees
16.95334, 8.03, ZMIN, ZTHICK
5.738, -5.725, XFMIN, XBMIN
1, # LAYERS
0.005, thickness of layer 1
0.521, 0.01, 1, 7,
MYLAR521ICRU
0.521, 0.01, 1, 23,
AIR521ICRU
0.521, 0.01, 1, 23,
AIR521ICRU
*********** start of CM JAWS with identifier JAWS ***********
20, RMAX
Secondary Collimators
2, # PAIRED BARS OR JAWS
Y
28, 35.8, 0.70000, 0.89500, -0.70000, -0.89500,
X
36.7, 44.5, 0.91750, 1.11250, -0.91750, -1.11250,
230
B.2. Example BEAMnrc input le
231
0.521, 0.01, 8, 7,
0.521, 0.01, 12, 12,
W521ICRU
0.521, 0.01, 13, 13,
W521ICRU
*********** start of CM DYNVMLC with identifier DYNVMLC ***********
25, RMAX
MLC based on 120 leaves millenium MLC Varian
1, 3, ORIENT, NGROUP
47.8, ZMIN
6.7, ZTHICK
0.533, 0.04, 0.04, 0.1354, 0.3676, 0.1396, 47.843, 48.126, 51.114, 51.325, 52.3873, 53.0573,
1.66, 54.1404, 54.405,
0.2492, 0.04, 0.04, 0.1054, 0.1336, 0.1471, 47.895, 48.1596, 49.0227, 49.4327, 1.7, 51.175,
51.177, 54.177, 54.296,
0.2338, 0.04, 0.04, 0.0754, 0.1405, 0.1316, 48.005, 48.124, 51.224, 51.325, 53.0673,
53.4773, 1.7, 54.1404, 54.405,
20, 1
80, 2
20, 1
-20, START
0.0057, LEAFGAP
0, ENDTYPE
8, ZFOCUS or RADIUS of leaf ends
0, ZFOCUS of leaf sides
-20, 20, 120
0.521, 0.01, 1, 23,
AIR521ICRU
0.521, 0.01, 1, 9, 0,
W521ICRU
0.521, 0.01, 1, 9,
W521ICRU
*********** start of CM SLABS with identifier AIRGAP ***********
30, RMAX
Airgap to patient
1, NSLABS
55.0889, ZMIN
44.911, 0.521, 0.01, 1, 10, 0
AIR521ICRU
*********************end of all CMs*****************************
:Start MC Transport Parameter:
Global ECUT= 0.521
Global PCUT= 0.01
Global SMAX= 1e10
ESTEPE= 0.25
XIMAX= 0.5
Boundary crossing algorithm= EXACT
Skin depth for BCA= 0
Electron-step algorithm= PRESTA-II
Spin effects= On
Brems angular sampling= Simple
Brems cross sections= BH
B.2. Example BEAMnrc input le
Bound Compton scattering= Off
Pair angular sampling= Simple
Photoelectron angular sampling= Off
Rayleigh scattering= On in regions
Atomic relaxations= Off
Electron impact ionization= Off
:Stop MC Transport Parameter:
232
B.3. Example DOSXYZnrc input le
B.3 Example DOSXYZnrc input le
50x50x50cm water phantom for ELL measurements #!GUI1.0
2
H2O521ICRU
AIR521ICRU
0.521, 0.01, 0, 0, 0
-3, -3, -5, 1
-15
14.5, 1
1, 1
14.5, 1
-15
14.5, 1
1, 1
14.5, 1
0
0.001, 10
0.01, 150
8.485, 1
0.01, 1
9.995, 1
0, 0, 0, 0, 0, 0, 0, 0
0, 0, 0, 0, 0, 0, 0, 0
0, 3, 0, 3, 0, 155, 0, 0
0, 0, 0, 0, 0, 0, 0, 0
2, 2, 0, 0, 0, 180, 0, 0, 180, 1, 20, 100, 100, 0
2, 2, 2, 0, 0, 0, 0, 0
/home/nick/egsnr cmp/BEAM TargetToPatient/5x5cm 100SSD 21.egsphsp1
2000000000, 0, 999, 657, 34, 100.0, 1, 0, 2, 0, , 0, 0, 0, 1, 1
:Start MC Transport Parameter:
Global ECUT= 0.521
Global PCUT= 0.01
Global SMAX= 1e10
ESTEPE= 0.25
XIMAX= 0.5
Boundary crossing algorithm= EXACT
Skin depth for BCA= 0
Electron-step algorithm= PRESTA-II
Spin effects= On
Brems angular sampling= Simple
Brems cross sections= BH
Bound Compton scattering= Off
Pair angular sampling= Simple
233
B.4. Comparison of Monte Carlo simulation with measured data
234
Photoelectron angular sampling= Off
Rayleigh scattering= Off
Atomic relaxations= Off
Electron impact ionization= Off
:Stop MC Transport Parameter:
B.4 Comparison of Monte Carlo simulation with
measured data
B.4. Comparison of Monte Carlo simulation with measured data
235
Figure B.2: Monte Carlo simulation data (MC) and Ion Chamber (IC) data for a
Varian 21EX linac at 1.5cm, 5cm and 10cm depths (a) X direction prole and (b) %
Depth Dose prole for a 5x5cm eld and (c) X direction prole and (d) % Depth
Dose prole for a 10x10cm eld
2
2
Appendix C
Statistical analysis
The two statistical analysis tools used in this thesis are Student's T-Test and the
Wilcoxon Rank Sum Test. Both of these tests are related to what is known as the
'normal' probability distribution. This is the most used probability distribution and
has a bell shape (the normal distribution is also known as the Gaussian distribution).
The shape of the normal distribution is symmetric about the mean of the distribution
and the width of the distribution is given by the standard deviation of the mean. The
standard deviation is the distance between the mean and the point of inection of the
curve on either side of the mean. The normal distribution, shown in Figure C.1 is a
probability distribution therefore the area under the curve is equal to 1. It must be
noted that the normal distribution is the distribution of the whole population, not
just a sample. Therefore, the range of X in Figure C.1 technically covers the range -1
and 1. The normal distribution is given by the following equation (Dawson & Trapp,
2004):
"
#
1
X 1
P (X ) =
(C.1)
2 2 To calculate probabilities using the normal distribution, the area under the curve
between the interval of interest is integrated. For example, if we want to know the
probability of a value falling between 2.2 and 2.4 in Figure C.1, equation C.1 is inte2
2
236
237
Figure C.1: The normal probability distribution for a mean of 2 and a standard
deviation of 0.5 shown for the interval of 0 to 4.
C.1. Student's t-test
238
grated with the limits of X being 2.2 and 2.4.
The standard normal distribution (or the z distribution) is a normal distribution
with a mean of 0 and a standard deviation of 1. For a given population, the probability
of obtaining a given result can be calculated using the standard normal distribution.
C.1 Student's t-test
Student's T-Test uses a variation of the normal probability distribution to calculate
the condence intervals for a given measurement. The t distribution is similar to that
of the normal distribution, however it represents the probability distribution of the
sample, not the whole population. For a given group, the condence intervals about
a mean can be calculated using the following (Dawson & Trapp, 2004):
Observed mean (condence coecient) (a measure of the variability of the
mean)
The condence coecient is known as the t value and is based on the level of
condence desired. The t value depends on the degrees of freedom (d.o.f) of a measurement, which is equal to N-1 where N is the number of measurements. The value of
t is calculated from the sample mean, the hypothesised population mean (as the actual
population mean is not known) and the standard error of the mean. In practice, to
obtain a given value of t, tables are used that contain the value of t for a given d.o.f
and desired condence level. For example, for a condence level of 95% with three
measurements (2 d.o.f), the value of t is equal to 4.303. The t distribution is similar
to the standard normal distribution however it is wider. As the d.o.f increases, the t
distribution approaches the standard normal distribution.
The measure of the variability of the mean is taken as the standard error of the
mean. That is, the standard deviation of the mean divided by the square root of the
number of measurements. Thus, for a given measurement, the condence intervals for
239
C.2. Wilcoxon rank sum test
the sample mean X with a sample standard deviation of SD for N measurements can
be calculated as (Dawson & Trapp, 2004):
SD
X t p
N
(C.2)
An example calculation of the 95% condence interval is as follows. Suppose we
had three measurements of the absorbed dose at a particular location - 205cGy, 201cGy
and 202cGy. The mean and standard deviation of these measurements is 202.67cGy
and 2.08cGy respectively. For a condence interval of 95%, and for 2 d.o.f, the value
of t is 4.303. Therefore we have the mean 95%CI as 202.67cGy 5.17cGy. Suppose we now have six measurements - 205cGy, 201cGy, 202cGy, 200cGy, 202cGy and
203cGy. The value of t for 95%CI and 5 d.o.f is 2.571 so the mean 95%CI becomes
202.17 2.56cGy. That is, the condence interval decreases in magnitude with more
measurements if they are normally distributed about the mean.
C.2 Wilcoxon rank sum test
The Wilcoxon Rank Sum Test is used to compare the mean between two independent
groups, to determine if the medians of the two groups are dierent (Dawson & Trapp,
2004).
The test can be described in the radiotherapy setting as follows. Two treatment
plans, Plan A and Plan B, are created for each patient in a given set of patients. A
certain parameter, say, the V25Gy parameter for an organ, is being compared between
the two groups. The Wilcoxon test compares the V25Gy parameter for each patient.
The absolute value of the dierences are ranked from smallest to largest - the smallest
dierence receives a rank of 1, the next smallest a rank of 2 and so on. The ranks in
each direction, that is, whether Plan A or Plan B has the higher V25Gy, are summed.
C.3. Spearman's rank correlation test
240
The smaller of these two sums is the test statistic, W. Basically, if a large dierence
occurs between Plan A and Plan B, one plan will have more patients with higher
ranks than the other plan, resulting in a higher value of W. If there is no signicant
dierence between the two plans, the low and high ranks will be evenly distributed
between the two plans and the value of W is lower. The value of W is then compared
to a table of all possible distributions of W values to obtain the value of p, which
describes the statistical signicance of the result (Dawson & Trapp, 2004).
The null hypothesis for this test is that the median dierence between pairs of
observations is zero. This diers from the t-test, where the null hypothesis is that the
mean dierence between pairs is zero (Dawson & Trapp, 2004).
C.3 Spearman's rank correlation test
Spearman's rank correlation test is used to test the relationship between two variables
in one group of subject. Correlation analysis is performed to determine if there is a
relationship between two values of the two variables (Altman & Gardner, 1988).
Spearman's rank correlation test converts each of the variables into ranks, with
the lowest values being ranked 1, the second lowest ranked 2 and so on. A correlation
analysis is then performed on the ranks McDonald (2008). In this analysis, a correlation coecient () is calculated for the ranks of the two sets of variables. The value of
will be between -1 and 1. For a perfect inverse relationship, =-1 and for a perfect
relationship, =1. When =0, there is no relationship between the two variables. In
other words, the correlation is stronger as approaces -1 or 1. The correlation coecient, , is then used with the degrees of freedom (number of pairs in the sample
minus 2) to nd the p value (using tables) which gives the statistical signicance of
the correlation. For example, if the value of results in a statistical signicance level
of less than 5% (p < 0.05), the probability of the relationship being due to chance is
C.3. Spearman's rank correlation test
5 in 100.
241
References
AACR, AIHW &. 2008. Cancer in Australia: An Overview, 2008. Tech. rept.
Afghan, M. K. N., Cao, D., Mehta, V., Wong, T., Ye, J., & Shepard, D. 2008. Volumetric modulated arc therapy for prostate cancer. Pages S665{S666 of: 50th Annual
Meeting of the American-Society-for-Therapeutic-Radiology-and Oncology.
Agostinelli, S. 2003. GEANT4-a simulation toolkit.
Nuclear Instruments & Methods
in Physics Research Section a-Accelerators Spectrometers Detectors and Associated
Equipment, 506(3), 250{303.
Ahnesjo, A. 1989. Collapsed Cone Convolution of Radiant Energy for Photon Dose
Calculation in Heterogeneous Media. Medical Physics, 16(4), 577{592.
Ahnesjo, A., & Aspradakis, M. M. 1999. Dose calculations for external photon beams
in radiotherapy. Physics in Medicine and Biology, 44(11), R99{R155.
Ahnesjo, A., Andreo, P., & Brahme, A. 1987. Calculation and Application of Point
Spread Functions for Treatment Planning with High-Energy Photon Beams. Acta
Oncologica, 26(1), 49{56.
Akazawa, C. 1989. Treatment of the scalp using photon and electron beams.
Dosim, 14(2), 129{31.
242
Med
References
243
Alber, M., & Nusslin, F. 1999. An objective function for radiation treatment optimization based on local biological measures. Physics In Medicine And Biology, 44(2),
479{493.
Altman, Douglas G, & Gardner, Martin J. 1988. Calculating condence intervals for
regression and correlation. British Medical Journal, 296, 1238{1242.
Arcangeli, S., Strigari, L., Soete, G., De Meerleer, G., Gomellini, S., Fonteyne, V.,
Storme, G., & Arcangeli, G. 2008. Clinical and Dosimetric Predictors of Acute
Toxicity after a 4-Week Hypofractionated External Beam Radiotherapy Regimen
for Prostate Cancer: Results from a Multicentric Prospective Trial. Int J Radiat
Oncol Biol Phys.
Arneld, M. R., Siantar, C. H., Siebers, J., Garmon, P., Cox, L., & Mohan, R. 2000a.
The impact of electron transport on the accuracy of computed dose. Medical Physics,
27(6), 1266{1274.
Arneld, M. R., Siebers, J. V., Kim, J. O., Wu, Q., Keall, P. J., & Mohan, R. 2000b. A
Method for Determining Multileaf Collimator Transmission and Scatter for Dynamic
Intensity Modulated Radiotherapy. Med Phys, 27, 2231{2241.
Assessment, Health Technology. 1997. NHS R&D HTA Programme; Diagnosis, management and screening of early localised prostatecancer. 1(2).
Baro, J., Sempau, J., Fernandezvarea, J. M., & Salvat, F. 1995. PENELOPE - An
Algorithm for Monte-Carlo Simulation of the Penetration and Energy-Loss of Electrons and Positrons in Matter. Nuclear Instruments & Methods in Physics Research
Section B-Beam Interactions with Materials and Atoms, 100(1), 31{46.
Bedford, J. L., Khoo, V. S., Oldham, M., Dearnaley, D. P., & Webb, S. 1999. A com-
References
244
parison of coplanar four-eld techniques for conformal radiotherapy of the prostate.
Radiother Oncol, 51(3), 225{35.
Bedford, J. L., Childs, P. J., Hansen, V. N., Warrington, A. P., Mendes, R. L., & Glees,
J. P. 2005. Treatment of extensive scalp lesions with segmental intensity-modulated
photon therapy. Int J Radiat Oncol Biol Phys, 62(5), 1549{58.
Bentley, R. E., & Milan, J. 1971. Interactive Digital Computer System For Radiotherapy Treatment Planning. British Journal Of Radiology, 44(527), 826{&.
Bentzen, S. M. 2006. Preventing or reducing late side eects of radiation therapy:
radiobiology meets molecular pathology. Nature Reviews Cancer, 6(9), 702{713.
Bentzen, Soren. 2009. Basic Clinical Radiobiology. 4th edn. A Hodder Arnold Publication. Chap. Dose-response relationships in radiotherapy, page 56067.
Bentzen, Soren M, & Tucker, Susan L. 1997. Quantifying the Position and Steepness
of Radiation Dose-Response Curves. Int J Radiat Oncol Biol Phys, 71, 531{542.
Boersma, L. J., van den Brink, M., Bruce, A. M., Shouman, T., Gras, L., te Velde,
A., & Lebesque, J. V. 1998. Estimation of the incidence of late bladder and rectum complications after high-dose (70-78 GY) conformal radiotherapy for prostate
cancer, using dose-volume histograms. Int J Radiat Oncol Biol Phys, 41(1), 83{92.
Bortfeld, T., & Webb, S. 2009. Single-Arc IMRT?
54(1), N9{N20.
Physics in Medicine and Biology,
Bortfeld, T., Schlegel, W., & Rhein, B. 1993. Decomposition of pencil beam kernels
for fast dose calculations in three-dimensional treatment planning. Medical Physics,
20, 311{318.
References
245
BORTFELD, T. R., KAHLER, D. L., WALDRON, T. J., & BOYER, A. L. 1994.
X-ray Field Compensation With Multileaf Collimators. International Journal of
Radiation Oncology Biology Physics, 28(3), 723{730.
Boswell, S., Tome, W., Jeraj, R., Jaradat, H., & Mackie, T. R. 2006. Automatic
registration of megavoltage to kilovoltage CT images in helical tomotherapy: an
evaluation of the setup verication process for the special case of a rigid head phantom. Med Phys, 33(11), 4395{404.
Boyer, A., & Mok, E. 1985. A photon dose distribution model employing convolution
calculations. Med Phys, 12(2), 169{77.
Boyer, A. L., & Li, S. 1997. Geometric Analysis of Light-eld Position of a Multileaf
Collimator with Curved ends. Med Phys, 24, 757{762.
Boyer, A. L., & Yu, C. X. 1999. Intensity-modulated radiation therapy with dynamic
multileaf collimators. Seminars in Radiation Oncology, 9(1), 48{59.
Bradley, J. D., Hope, A., & El Naqa, I. et al. 2007. A nomogram to predict radiation
pneumonitis, derived from a combined analysis of RTOG 9311 and institutional
data. Int J Radiat Oncol Biol Phys, 69, 985{992.
Brahme, A. 1984. Dosimetric Precission Requirements in Radiation Therapy.
Radiol Oncol, 23, 379{391.
Acta
Brahme, A Roos, J-E Lax I. 1982. Solution of an integral equation encountered in
radiation therapy. Phys Med Biol, 27, 1221{9.
Brenner, D. J., & Hall, E. J. 1999. Fractionation and protraction for radiotherapy of
prostate carcinoma. Int J Radiat Oncol Biol Phys, 43(5), 1095{101.
References
246
Brenner, D. J., Martinez, A. A., Edmundson, G. K., Mitchell, C., Thames, H. D., &
Armour, E. P. 2002. Direct evidence that prostate tumors show high sensitivity to
fractionation (low alpha/beta ratio), similar to late-responding normal tissue. Int J
Radiat Oncol Biol Phys, 52(1), 6{13.
Briesmeister, J F. 1986. MCNP - A General Monte Carlo Code for Neutron and
Photon Transport. Tech. rept. LA-7396-M.
Butson, M. J., Rozenfeld, A., Mathur, J. N., Carolan, M., Wong, T. P. Y., & Metcalfe,
P. E. 1996. A new radiotherapy surface dose detector: The MOSFET. Medical
Physics, 23(5), 655{658.
Bylund, K. C., Bayouth, J. E., Smith, M. C., Hass, A. C., Bhatia, S. K., & Buatti, J. M. 2008. Analysis of interfraction prostate motion using megavoltage cone
beam computed tomography. International Journal Of Radiation Oncology Biology
Physics, 72(3), 949{956.
Bzdusek, Karl, Friberger, Henrick, Eriksson, Kjell, Hardemark, Bjorn, Robinson,
David, & Kaus, Michael. 2009. Development and Evaluation of an Ecient Approach to Volumetric Arc Therapy Planning. Medical Physics, 36(6), 2328{2339.
Cadman, P., Bassalow, R., Sidhu, N. P. S., Ibbott, G., & Nelson, A. 2002. Dosimetric considerations for validation of a sequential IMRT process with a commercial
treatment planning system. Physics in Medicine and Biology, 47(16), 3001{3010.
Cadman, Patrick, McNutt, Todd, & Bzdusek, Karl. 2005. Validation of Physics Improvements for IMRT with a Commercial Treatment-planning System. J Appl Clin
Med Phys, 6, 74{86.
Carrasco, P., Jornet, N., Duch, M. A., Weber, L., Ginjaume, M., Eudaldo, T., Jurado,
D., Ruiz, A., & Ribas, M. 2004. Comparison of dose calculation algorithms in
References
247
phantoms with lung equivalent heterogeneities under conditions of lateral electronic
disequilibrium. Medical Physics, 31(10), 2899{2911.
Chapet, O., Thomas, E., Kessler, M. L., Fraass, B. A., & Ten Haken, R. K. 2005.
Esophagus sparing with IMRT in lung tumor irradiation: an EUD-based optimization technique. Int J Radiat Oncol Biol Phys, 63(1), 179{87.
Chappell, R., Fowler, J., & Ritter, M. 2004. New data on the value of alpha/beta{
evidence mounts that it is low. Int J Radiat Oncol Biol Phys, 60(3), 1002{3.
Cheek, D., Hogstrom, K., Gibbons, J., & Rosen, I. 2007. SU-FF-T-213: Evaluation of
Dose From TomoTherapy Irradiation of Supercial PTVs. Medical Physics, 34(6),
2450.
Chen, M. E., Johnston, D. A., Tang, K., Babaian, R. J., & Troncoso, P. 2000. Detailed
mapping of prostate carcinoma foci: biopsy strategy implications. Cancer, 89(8),
1800{9.
Cherpak, A., Studinski, R. C., & Cygler, J. E. 2007. MOSFET detectors in quality
assurance of tomotherapy treatments. Radiother Oncol.
Cheung, T., Butson, M. J., & Yu, P. K. 2005. Post-irradiation colouration of
Gafchromic EBT radiochromic lm. Phys Med Biol, 50(20), N281{5.
Choi, B., & Deasy, J. O. 2002. The generalized equivalent uniform dose function as a
basis for intensity-modulated treatment planning. Phys Med Biol, 47(20), 3579{89.
Chow, James C. L., & Grigorov, Grigor N. 2008. Surface dosimetry for oblique tangential photon beams: A Monte Carlo simulation study. Medical Physics, 35(1),
70{76.
References
248
Chvetsov, A. V., Dempsey, J. F., & Palta, J. R. 2007. Optimization of equivalent
uniform dose using the L-curve criterion. Phys Med Biol, 52(19), 5973{84.
CONVERY, D. J., & ROSENBLOOM, M. E. 1992. The Generation of Intensitymodulated Fields For Conformal Radiotherapy By Dynamic Collimation. Physics
In Medicine and Biology, 37(6), 1359{1374.
Court, L. E., Dong, L., Lee, A. K., Cheung, R., Bonnen, M. D., O'Daniel, J., Wang, H.,
Mohan, R., & Kuban, D. 2005. An automatic CT-guided adaptive radiation therapy
technique by online modication of multileaf collimator leaf positions for prostate
cancer. International Journal Of Radiation Oncology Biology Physics, 62(1), 154{
163.
Cozzarini, C., Fiorino, C., Di Muzio, N., Alongi, F., Broggi, S., Cattaneo, M., Montorsi, F., Rigatti, P., Calandrino, R., & Fazio, F. 2007. Signicant reduction of
acute toxicity following pelvic irradiation with helical tomotherapy in patients with
localized prostate cancer. Radiother Oncol, 84(2), 164{70.
D'Amico, A. V., Manola, J., McMahon, E., Loredo, M., Lopes, L., Ching, J., Albert,
M., Hurwitz, M., Suh, W. W., Vivenzio, T. A., & Beard, C. 2006. A prospective evaluation of rectal bleeding after dose-escalated three-dimensional conformal
radiation therapy using an intrarectal balloon for prostate gland localization and
immobilization. Urology, 67(4), 780{4.
Davies, A, Foo, K, Miller, A, Arnold, A., Bailey, M., Carolan, M., Dixon, J., Fox,
C., Nasser, E., & Williams, M. 2008a. Advantage of single-phase, uniform-margin
prostate radiotherapy. Australasian Radiology, 52, A111.
Davies, A, Foo, K, Miller, A, Arnold, A., Bailey, M., Carolan, M., Dixon, J., Fox,
249
References
C., Nasser, E., & Williams, M. 2008b.
prostate radiotherapy.
Advantage of single-phase, uniform-margin
Dawson, Beth, & Trapp, Rober G. 2004.
McGraw-Hill.
Basic & Clinical Biostatistics.
4 edn.
Dearnaley, D. P., Sydes, M. R., Graham, J. D., Aird, E. G., Bottomley, D., Cowan,
R. A., Huddart, R. A., Jose, C. C., Matthews, J. H., Millar, J., Moore, A. R.,
Morgan, R. C., Russell, J. M., Scrase, C. D., Stephens, R. J., Syndikus, I., &
Parmar, M. K. 2007. Escalated-dose versus standard-dose conformal radiotherapy
in prostate cancer: rst results from the MRC RT01 randomised controlled trial.
Lancet Oncol, 8(6), 475{87.
Deasy, J. O., Blanco, A. I., & Clark, V. H. 2003. CERR: A computational environment
for radiotherapy research. Medical Physics, 30(5), 979{985.
Devic, S., Seuntjens, J., Abdel-Rahman, W., Evans, M., Olivares, M., Podgorsak,
E. B., Vuong, T., & Soares, C. G. 2006. Accurate skin dose measurements using
radiochromic lm in clinical applications. Medical Physics, 33(4), 1116{1124.
Dogan, N., Leybovich, L. B., Sethi, A., & Emami, B. 2003. Automatic feathering of
split elds for step-and-shoot intensity modulated radiation therapy. Phys. Med.
Biol., 48, 1133{1140.
Emami, B., Lyman, J., Brown, A., Coia, L., Goitein, M., Munzenrider, J. E., Shank,
B., Solin, L. J., & Wesson, M. 1991. Tolerance of Normal Tissue to Therapeutic
Irradiation. International Journal of Radiation Oncology Biology Physics, 21(1),
109{122.
Fenwick, J. D., Khoo, V. S., Nahum, A. E., Sanchez-Nieto, B., & Dearnaley, D. P. 2001.
Correlations between dose-surface histograms and the incidence of long-term rectal
References
250
bleeding following conformal or conventional radiotherapy treatment of prostate
cancer. Int J Radiat Oncol Biol Phys, 49(2), 473{80.
Fiorino, C., Broggi, S., Corletto, D., Cattaneo, G. M., & Calandrino, R. 2000. Conformal irradiation of concave-shaped PTVs in the treatment of prostate cancer by
simple 1D intensity-modulated beams. Radiother Oncol, 55(1), 49{58.
Fiorino, C., Cozzarini, C., Vavassori, V., Sanguineti, G., Bianchi, C., Cattaneo, G. M.,
Foppiano, F., Magli, A., & Piazzolla, A. 2002. Relationships between DVHs and
late rectal bleeding after radiotherapy for prostate cancer: analysis of a large group
of patients pooled from three institutions. Radiother Oncol, 64(1), 1{12.
Fiorino, C., Foppiano, F., Franzone, P., Broggi, S., Castellone, P., Marcenaro, M.,
Calandrino, R., & Sanguineti, G. 2005. Rectal and bladder motion during conformal
radiotherapy after radical prostatectomy. Radiother Oncol, 74(2), 187{95.
Fiorino, Claudio, Fellin, Gianni, Rancati, Tiziana, Vavassori, Vittorio, Bianchi, Carla,
Borca, Valeria Casanova, Girelli, Giuseppe, Mapelli, Marco, Menegotti, Loris, Nava,
Simona, & Valdagni, Riccardo. 2008. Clinical and Dosimetric Predictors of Late
Rectal Syndrome After 3D-CRT for Localized Prostate Cancer: Preliminary Results of a Multicenter Prospective Study. International Journal of Radiation OncologyBiologyPhysics, 70(4), 1130{1137.
Fogliata, A., Vanetti, E., Albers, D., Brink, C., Clivio, A., Knoos, T., Nicolini, G., &
Cozzi, L. 2007. On the dosimetric behaviour of photon dose calculation algorithms
in the presence of simple geometric heterogeneities: comparison with Monte Carlo
calculations. Physics in Medicine and Biology, 52(5), 1363{1385.
Fowler, J., Chappell, R., & Ritter, M. 2001. Is alpha/beta for prostate tumors really
low? Int J Radiat Oncol Biol Phys, 50(4), 1021{31.
References
251
Fowler, J. F. 1989. The Linear-Quadratic Formula and Progress in Fractionated Radiotherapy. British Journal of Radiology, 62(740), 679{694.
Fowler, J. F., Nahum, A. E., & Orton, C. G. 2006. The best radiotherapy for the
treatment of prostate cancer involves hypofractionation. Medical Physics, 33(9),
3081{3084.
Goldner, G., Geinitz, H., Wachter, S., Becker, G., Zimmermann, F., Wachter-Gerstner,
N., Glocker, S., Potzi, R., Wambersie, A., Bamberg, M., Molls, M., Feldmann, H.,
& Potter, R. 2006. 3-D Conformal radiotherapy of localized prostate cancer within
an Austrian-German multicenter trial: a prospective study of patients' acceptance
of the rectal balloon during treatment. Wiener Klinische Wochenschrift, 118(7-8),
224{229.
Gospodarowicz, M.K., Miller, D., Groome, P.A., Greene, F.L., Logan, P.A., & Sobin,
L.H. 2004. The process for continuous improvement of the TNM classication.
Cancer, 100, 1{5.
Guckenberger, M Richter, A Krieger T Wilbert J Baier K Flentje M. 2009. Is a single arc sucient in volumetric-modulated arc therapy (VMAT) for complex-shaped
target volumes? Radiotherapy and Oncology, 93, 259{265.
Gulliford, S.L., Foo, K., Morgan, R. C., Aird, E. G., Bidmead, A. M., Critchley, H.,
Evans, P.M., Gianolini, S, Mayles, W. P., Moore, A. R., Sanchez, B., Partridge, M.,
Sydes, M. R., Webb, S., & Dearnaley, D. 2009. Dose-volume constraints to reduce
rectal side eects from prostate radiotherapy: Evidencefrom the MRC RT01 trial
ISRCTN 47772397. Int J Radiat Oncol Biol Phys, In Press.
Hall, Eric J., & Wuu, Cheng-Shie. 2003. Radiation-induced second cancers:
References
the impact of 3D-CRT and IMRT.
ogyBiologyPhysics, 56(1), 83{88.
252
International Journal of Radiation Oncol-
Hanks, G. E., Hanlon, A. L., Schultheiss, T. E., Pinover, W. H., Movsas, B., Epstein,
B. E., & Hunt, M. A. 1998. Dose escalation with 3D conformal treatment: ve year
outcomes, treatment optimization, and future directions. Int J Radiat Oncol Biol
Phys, 41(3), 501{10.
Hardemark, Bjorn, Liander, Anders, Rehbinder, Henrik, & Lof, Johan. 2003. Direct
Machine Parameter Optimization with RayMachine in Pinnacle. RaySearch Laboratories.
Heath, Emily, & Seuntjens, Jan. 2003. Development and Validation of a BEAMnrc
Component Module for Accurate Monte Carlo Modelling of the Varian Dynamic
Millennium Multileaf Collimator. Phys Med Biol, 48, 4045{4063.
Higgins, P. D., Han, E. Y., Yuan, J. L., Hui, S., & Lee, C. K. 2007. Evaluation
of surface and supercial dose for head and neck treatments using conventional or
intensity-modulated techniques. Phys Med Biol, 52(4), 1135{46.
Hoban, P. W., Murray, D. C., Metcalfe, P. E., & Round, W. H. 1990. Superposition
dose calculation in lung for 10MV photons. Australas Phys Eng Sci Med, 13(2),
81{92.
Holmes-Siedle, Andrew. 1974. The Space Charge Dosimeter. Nuclear Instrumentation
and Methods, 121, 169{179.
Howard, A., & Pelc, S.R. 1953. Synthesis of deoxyribonucleic acid innormal and
irradiated cells and its relation to chromosome breakage. Heredity, 6, 261{273.
Hsi, R. A., Vali, F., Parsai, J., Garver, E., Madsen, B., Pham, H., Song, G., Badiozamani, K., & Cho, P. 2008. Evaluation of interfraction and intrafraction prostate
253
References
motion during the treatment of prostate cancer using the Calypso 4D localization
system. International Journal Of Radiation Oncology Biology Physics, 72(1), S300{
S300.
Huang, E., Dong, L., Chandra, A., Kuban, D. A., Rosen, I. I., Evans, A., & Pollack, A.
2002a. Intrafraction prostate motion during IMRT for prostate cancer. International
Journal Of Radiation Oncology Biology Physics, 53(2), 261{268.
Huang, E. H., Pollack, A., Levy, L., Starkschall, G., Dong, L., Rosen, I., & Kuban,
D. A. 2002b. Late rectal toxicity: dose-volume eects of conformal radiotherapy for
prostate cancer. Int J Radiat Oncol Biol Phys, 54(5), 1314{21.
ICRP. 1991. The biological basis for dose limitation in the skin. A report of a Task
Group of Committee 1 of the International Commission on Radiological Protection.
Ann ICRP, 22(2), 1{104.
IEC. 1996. Radiotherapy equipment - Coordinates, movements and scales.
Institute, NSW Cancer. 2007.
Cancer Instite.
Prostate Cancer and PSA Testing.
Tech. rept. NSW
ISP. 2007. Gafchromic EBT: Self Developing Film for Radiotherapy Dosimetry. White
Paper.
ISP. 2009. Gafchromic EBT2: Self Developing Film for Radiotherapy Dosimetry. Tech.
rept.
Jackson, A., Skwarchuk, M. W., Zelefsky, M. J., Cowen, D. M., Venkatraman, E. S.,
Levegrun, S., Burman, C. M., Kutcher, G. J., Fuks, Z., Liebel, S. A., & Ling,
C. C. 2001. Late rectal bleeding after conformal radiotherapy of prostate cancer. II.
Volume eects and dose-volume histograms. Int J Radiat Oncol Biol Phys, 49(3),
685{98.
References
Jeraj, R., & Keall, P. 1999. Monte Carlo-based inverse treatment planning.
in Medicine and Biology, 44(8), 1885{1896.
254
Physics
Joiner, M. C., Marples, B., Lambin, P., Short, S. C., & Turesson, I. 2000. Low-dose
hypersensitivity: Current status and possible mechanisms. Pages 379{389 of: 1st
International Conference on Translational Research and Pre-Clinical Strategies in
Clinical Radio-Oncology (ICTR 2000).
Jones, A. O., & Das, I. J. 2005. Comparison of inhomogeneity correction algorithms
in small photon elds. Medical Physics, 32(3), 766{776.
Jones, L. 1999. Radiotherapy Optimisation: Biological and Physical Concepts. Ph.D.
thesis, University of New South Wales.
Kallman, P., Agren, A., & Brahme, A. 1992. Tumor and Normal Tissue Responses
to Fractionated Nonuniform Dose Delivery. International Journal of Radiation Biology, 62(2), 249{262. ISI Document Delivery No.: JL581 Times Cited: 199 Cited
Reference Count: 36.
Kawrakow, I., & Fippel, M. 2000. VMC++, a fast MC algorithm for radiation treatment planning. Use Of Computers In Radiation Therapy, 126{128.
Kawrakow, I., Rogers, D. W. O., & Walters, B. R. B. 2004. Large eciency improvements in BEAMnrc using directional bremsstrahlung splitting. Medical Physics,
31(10), 2883{2898.
Keall, P. J., & Hoban, P. W. 1996. Superposition dose calculation incorporating Monte
Carlo generated electron track kernels. Med Phys, 23(4), 479{85.
Kennedy, J.M. 2005. High-intensity focused ultrasound in the treatment of solid tumours. Nature Reviews Cancer, 5(4), 321{327.
References
255
Khoo, V. S., Bedford, J. L., Webb, S., & Dearnaley, D. P. 2000. An evaluation of
three-eld coplanar plans for conformal radiotherapy of prostate cancer. Radiother
Oncol, 55(1), 31{40.
Khuntia, D., Jaradat, H., Orton, N., Tome, W., Mehta, M. P., & Welsh, J. S. 2006.
Helical tomotherapy as a means of administering total or partial scalp irradiation:
In regards to Bedford et al. (Int J Radiat Oncol Biol Phys 2005;62:1549-1558). Int
J Radiat Oncol Biol Phys, 64(4), 1288{9; author reply 1289{90.
King, C. R., & Fowler, J. F. 2001. A simple analytic derivation suggests that prostate
cancer alpha/beta ratio is low. Int J Radiat Oncol Biol Phys, 51(1), 213{4.
Kirkby, C., Stanescu, T., Rathee, S., Carlone, M., Murray, B., & Fallone, B. G. 2008.
Patient dosimetry for hybrid MRI-radiotherapy systems. Medical Physics, 35(3),
1019{1027.
Kissick, M. W., Fenwick, J., James, J. A., Jerai, R., Kapatoes, J. M., Keller, H.,
Mackie, T. R., Olivera, G., & Soisson, E. T. 2005. The helical tomotherapy thread
eect. Medical Physics, 32(5), 1414{1423.
Kjaer-Kristoersen, F., Ohlhues, L., Medin, J., & Korreman, S. 2008. RapidArc
volumetric modulated therapy planning for prostate cancer patients. Pages 227{232
of: Symposium of the Nordic-Association-of-Clinical-Physics. Kjaer-Kristoersen,
Flemming Ohlhues, Lars Medin, Joakim Korreman, Stine.
Kornelsen, R. O., & Young, M. E. 1982. Changes in the dose-prole of a 10 MV x-ray
beam within and beyond low density material. Med Phys, 9(1), 114{6.
Kuban, D. A., Tucker, S. L., Dong, L., Starkschall, G., Huang, E. E., Cheung, M. R.,
Lee, A. K., & Pollack, A. 2008. Long-term results of the M. D. Anderson random-
References
ized dose-escalation trial for prostate cancer.
Oncology Biology Physics, 70(1), 67{74.
256
International Journal of Radiation
Kung, J. H., & Chen, G. T. 2000. Intensity Modulated Radiotherapy dose Delivery
Error from Radiation Field Oset Inaccuracy. Med Phys, 27, 1617{1622.
Kupelian, P. A., Reddy, C. A., Klein, E. A., & Willoughby, T. R. 2001. Shortcourse intensity-modulated radiotherapy (70 GY at 2.5 GY per fraction) for localized
prostate cancer: preliminary results on late toxicity and quality of life. Int J Radiat
Oncol Biol Phys, 51(4), 988{93.
Kupelian, P. A., Reddy, C. A., Carlson, T. P., & Willoughby, T. R. 2002a. Dose/volume relationship of late rectal bleeding after external beam radiotherapy for localized
prostate cancer: Absolute or relative rectal volume? Cancer Journal, 8(1), 62{66.
Kupelian, P. A., Reddy, C. A., Carlson, T. P., Altsman, K. A., & Willoughby, T. R.
2002b. Preliminary observations on biochemical relapse-free survival rates after
short-course intensity-modulated radiotherapy (70 Gy at 2.5 Gy/fraction) for localized prostate cancer. International Journal of Radiation Oncology Biology Physics,
53(4), 904{912.
Kupelian, P. A., Willoughby, T. R., Reddy, C. A., Klein, E. A., & Mahadevan, A. 2007.
Hypofractionated intensity-modulated radiotherapy (70 Gy at 2.5 Gy per fraction)
for localized prostate cancer: Cleveland Clinic experience. Int J Radiat Oncol Biol
Phys, 68(5), 1424{30.
Kutcher, G. J., & Burman, C. 1989. Calculation of Complication Probability Factors for Non-Uniform Normal Tissue Irradiation - The Eective Volume Method.
International Journal of Radiation Oncology Biology Physics, 16(6), 1623{1630.
References
257
Kwan, I. S., Rosenfeld, A. B., Qi, Z. Y., Wilkinson, D., Lerch, M. L. F., Cutajar, D. L.,
Safavi-Naeni, M., Butson, M., Bucci, J. A., Chin, Y., & Perevertaylo, V. L. 2007.
Skin dosimetry with new MOSFET detectors. Pages 929{932 of: 15th International
Conference on Solid State Dosimetry.
Kwan, I. S., Lee, B., Yoo, A.J., Cho, D., Jang, K., Shin, S., Carolan, M., Lerch, M.,
Perevertaylo, V. L., & Rosenfeld, A.B. 2008a. Comparison of the New MOSkin
Detector and Fiber Optic Dosimetry System for Radiotherapy. Journal of Nuclear
Science and Technology, Japan S5(June), 518{521.
Kwan, I. S., Howie, A., Lerch, M., Lee, B., chin, Y.S., Bucci, J., Perevertaylo, V. L.,
& Rosenfeld, A. B. 2008b. Measurement of Rectal Dose during HDR Brachytherapy
using the new MOSkin Dosimeter. Journal of Nuclear Science and Technology,
Japan S5(June), 481{484.
Kwan, I. S., Wilkinson, D., Cutajar, D., Lerch, M., Rosenfeld, A., Howie, A., Bucci,
J., Chin, Y., & Perevertaylo, V. L. 2009. The eect of rectal heterogeneity on wall
dose in high dose rate brachytherapy. Medical Physics, 36(1), 224{232.
Lagendijk, J. J. W., Raaymakers, B. W., Raaijmakers, A. J. E., Overweg, J., Brown,
K. J., Kerkhof, E. M., van der Put, R. W., Hardemark, B., van Vutpen, M., &
van der Heide, U. A. 2008. MRI/linac integration. Radiotherapy And Oncology,
86(1), 25{29.
Leborgne, F., & Fowler, J. 2008. Acute Toxicity After Hypofractionated Conformal
Radiotherapy for Localized Prostate Cancer: Nonrandomized Contemporary Comparison with Standard Fractionation. Int J Radiat Oncol Biol Phys.
Lee, N., Chuang, C., Quivey, J. M., Phillips, T. L., Akazawa, P., Verhey, L. J., & Xia,
References
258
P. 2002. Skin toxicity due to intensity-modulated radiotherapy for head-and-neck
carcinoma. Int J Radiat Oncol Biol Phys, 53(3), 630{7.
Li, X. A., Yu, C., & Holmes, T. 2000. A systematic evaluation of air cavity dose
perturbation in megavoltage x-ray beams. Med Phys, 27(5), 1011{7.
Li, X. A., Qi, X. S., Pitterle, M., Kalakota, K., Mueller, K., Erickson, B. A., Wang,
D., Schultz, C. J., Firat, S. Y., & Wilson, J. F. 2007. Interfractional variations in
patient setup and anatomic change assessed by daily computed tomography. Int J
Radiat Oncol Biol Phys, 68(2), 581{91.
Lian, Cheryl. 2009. MOSkin Filtering with Copper - A Theoretical Investigation. Personal Communication.
Lin, S. H., Latronico, D., Teslow, T., & Bajaj, G. K. 2008. A highly reproducible
bolus immobilization technique for the treatment of scalp malignancies. Med Dosim,
33(1), 30{5.
Ling, C. C., Zhang, P., Archambault, Y., Bocanek, J., Tang, G., & LoSasso, T. 2008.
Commissioning and quality assurance of RapidArc radiotherapy delivery system.
International Journal of Radiation Oncology Biology Physics, 72(2), 575{581.
Livsey, J. E., Cowan, R. A., Wylie, J. P., Swindell, R., Read, G., Khoo, V. S., &
Logue, J. P. 2003. Hypofractionated conformal radiotherapy in carcinoma of the
prostate: ve-year outcome analysis. Int J Radiat Oncol Biol Phys, 57(5), 1254{9.
Locke, J., Low, D. A., Grigireit, T., & Chao, K. S. 2002. Potential of tomotherapy for
total scalp treatment. Int J Radiat Oncol Biol Phys, 52(2), 553{9.
Logue, J. P., Cowan, R. A., & Hendry, J. H. 2001. Hypofractionation for prostate
cancer. Int J Radiat Oncol Biol Phys, 49(5), 1522{3.
References
259
LoSasso, T., Chui, C. S., & Ling, C. C. 1998. Physical and Dosimetric Aspects of a
Multileaf Collimation System used in the Dynamic mode for Implementing Intensity
Modulated Radiotherapy. Med Phys, 25, 1919{1927.
Ludlum, E., Mu, G. W., Weinberg, V., Roach, M., Verhey, L. J., & Xia, P. 2007.
An algorithm for shifting MLC shapes to adjust for daily prostate movement during
concurrent treatment with pelvic lymph nodes. Medical Physics, 34(12), 4750{4756.
Luxton, G., Hancock, S. L., & Boyer, A. L. 2004. Dosimetry and radiobiologic model
comparison of IMRT and 3D conformal radiotherapy in treatment of carcinoma of
the prostate. International Journal of Radiation Oncology Biology Physics, 59(1),
267{284.
Lyman, J. T. 1985. Complication Probability as Assessed from Dose Volume Histograms. Radiation Research, 104(2), S13{S19.
Ma, L. J., Boyer, A. L., Xing, L., & Ma, C. M. 1998. An optimized leaf-setting algorithm for beam intensity modulation using dynamic multileaf collimators. Physics
in Medicine and Biology, 43(6), 1629{1643.
MacKay, R. I., Hendry, J. H., Moore, C. J., Williams, P. C., & Read, G. 1997. Predicting late rectal complications following prostate conformal radiotherapy using
biologically eective doses and normalized dose-surface histograms. Br J Radiol,
70(833), 517{26.
MacKenzie, M. A., Lachaine, M., Murray, B., Fallone, B. G., Robinson, D., & Field,
G. C. 2002. Dosimetric Verication of Inverse Planned step and Shoot Multileaf
Collimator Fields from a Commercial Treatment Planning System. J Appl Clin
Med Phys, 3, 97{9109.
References
260
Mackie, T. R., Scrimger, J. W., & Battista, J. J. 1985a. A convolution method of
calculating dose for 15-MV x rays. Med Phys, 12(2), 188{96.
Mackie, T. R., el Khatib, E., Battista, J., Scrimger, J., Van Dyk, J., & Cunningham,
J. R. 1985b. Lung dose corrections for 6- and 15-MV x rays. Med Phys, 12(3),
327{32.
Mackie, T. R., Bielajew, A. F., Rogers, D. W., & Battista, J. J. 1988. Generation of
photon energy deposition kernels using the EGS Monte Carlo code. Phys Med Biol,
33(1), 1{20.
Martens, C., Reynaert, N., De Wagter, C., Nilsson, P., Coghe, M., Palmans, H.,
Thierens, H., & De Neve, W. 2002. Underdosage of the upper-airway mucosa for
small elds as used in intensity-modulated radiation therapy: a comparison between
radiochromic lm measurements, Monte Carlo simulations, and collapsed cone convolution calculations. Med Phys, 29(7), 1528{35.
Marzi, S., Arcangeli, G., Saracino, B., Petrongari, M. G., Bruzzaniti, V., Iaccarino,
G., Landoni, V., Soriani, A., & Benassi, M. 2007. Relationships between rectal
wall dose-volume constraints and radiobiologic indices of toxicity for patients with
prostate cancer. Int J Radiat Oncol Biol Phys, 68(1), 41{9.
Massey, J. B. 1962. Dose distribution problems in megavoltage therapy. I. The problem
of air spaces. Br J Radiol, 35, 736{8.
McBain, Catherine A., Khoo, Vincent S., Buckley, David L., Sykes, Jonathan S.,
Green, Melanie M., Cowan, Richard A., Hutchinson, Charles E., Moore, Christopher J., & Price, Patricia M. 2009. Assessment of Bladder Motion for Clinical Radiotherapy Practice Using Cine-Magnetic Resonance Imaging. International Journal
of Radiation OncologyBiologyPhysics, In Press, Corrected Proof.
References
261
McDonald, J.H. 2008. Handbook of Biological Statistics. Sparky House Publishing,
Baltimore, Maryland.
McGary, J. E., Teh, B. S., Butler, E. B., & Grant, W., 3rd. 2002. Prostate immobilization using a rectal balloon. J Appl Clin Med Phys, 3(1), 6{11.
McNutt, Todd. 2002. Dose calculations: collapsed cone convolution superposition and
delta pexel beam. Tech. rept. Philips White Paper No. 4535 983 02474.
Mehta, M., Hoban, P., & Mackie, T. R. 2009. Commissioning and Quality Assurance
of Rapidarc Radiotherapy Delivery System: In Regard To Ling Et Al. (int J Radiat
Oncol Biol Phys 2008;72;575-581): Absence of Data Does Not Constitute Proof;
the Proof Is In Tasting the Pudding. International Journal of Radiation Oncology
Biology Physics, 75(1), 4{6.
Mestrovic, A., Milette, M. P., Nichol, A., Clark, B. G., & Otto, K. 2007. Direct
aperture optimization for online adaptive radiation therapy. Medical Physics, 34(5),
1631{1646.
Metcalfe, P., Tangboonduangjit, P., & White, P. 2004. Intensity-modulated Radiation
Therapy: Overlapping Co-axial Modulated Fields. Phys Med Biol, 49, 3629{3637.
Metcalfe, P. E., & Battista, J. J. 1988. Accuracy of inhomogeneity corrections in lung
irradiated with high energy X-rays. Australas Phys Eng Sci Med, 11(2), 67{75.
Metcalfe, P. E., Wong, T. P., & Hoban, P. W. 1993. Radiotherapy X-ray beam
inhomogeneity corrections: the problem of lateral electronic disequilibrium in lung.
Australas Phys Eng Sci Med, 16(4), 155{67.
Metcalfe, Peter, Kron, Tomas, & Hoban, Peter. 2007. The Physics of Radiotherapy
X-Rays and Electrons. Madison, WI: Medical Physics Publishing.
References
262
Milas, L., & Hittelman, W. N. 2009. Cancer Stem Cells and Tumor Response to
Therapy: Current Problems and Future Prospects. Seminars in Radiation Oncology,
19(2), 96{105.
Mohan, R., Chui, C., & Lidofsky, L. 1986. Dierential Pencil Beam Dose Computation
Model for Photons. Medical Physics, 13(1), 64{73.
Moiseenko, V., Deasy, J., & Van Dyke, J. 2005. The Modern Technology of Radiation
Oncology. Vol. 2. Medical Physics Publishing. Chap. Radiobiological Modeling for
Treatment Planning, pages 185{214.
Morgan, W. F. 2003. Non-targeted and delayed eects of exposure to ionizing radiation: I. Radiation-induced genomic instability and bystander eects in vitro.
Radiation Research, 159(5), 567{580.
Nagasawa, H., & Little, J. B. 1992. Induction of Sister Chromated Exchanges by
Extremely Low-Doses of Alpha-Particles. Cancer Research, 52(22), 6394{6396.
Nahum, A. E., Movsas, B., Horwitz, E. M., Stobbe, C. C., & Chapman, J. D. 2003.
Incorporating clinical measurements of hypoxia into tumor local control modeling
of prostate cancer: Implications for the alpha/beta ratio. International Journal Of
Radiation Oncology Biology Physics, 57(2), 391{401.
Namiki, S., Ishidoya, S., Tochigi, T., Kawamura, S., Kuwahara, M., Terai, A.,
Yoshimura, K., Numata, I., Satoh, M., Saito, S., Takai, Y., Yamada, S., & Arai, Y.
2006. Health-related quality of life after intensity modulated radiation therapy for
localized prostate cancer: Comparison with conventional and conformal radiotherapy. Japanese Journal of Clinical Oncology, 36(4), 224{230.
Niemierko, A. 1997. Reporting and analyzing dose distributions: a concept of equivalent uniform dose. Med Phys, 24(1), 103{10.
263
References
Nilsson, B., & Schnell, P. O. 1976. Build-up eects at air cavities measured with thin
thermoluminescent dosimeters. Acta Radiol Ther Phys Biol, 15(5), 427{32.
Niroomand-Rad, A., Blackwell, C. R., Coursey, B. M., Gall, K. P., Galvin, J. M.,
McLaughlin, W. L., Meigooni, A. S., Nath, R., Rodgers, J. E., & Soares, C. G.
1998. Radiochromic lm dosimetry: Recommendations of AAPM Radiation Therapy Committee Task Group 55. Medical Physics, 25(11), 2093{2115.
Noel, C., Parikh, P. J., Roy, M., Kupelian, P., Mahadevan, A., Weinstein, G., Enke,
C., Flores, N., Beyer, D., & Levine, L. 2009. Prediction Of Intrafraction Prostate
Motion: Accuracy Of Pre- And Post-Treatment Imaging And Intermittent Imaging.
International Journal Of Radiation Oncology Biology Physics, 73(3), 692{698.
Oborn, Brad. 2008 (September).
Communication.
High Resolution Surface Dose Simulations.
Private
Olafsson, A., Jeraj, R., & Wright, S. J. 2005. Optimization of intensity-modulated
radiation therapy with biological objectives. Phys Med Biol, 50(22), 5357{79.
Orton, N., Jaradat, H., Welsh, J., & Tome, W. 2005. Total scalp irradiation using
helical tomotherapy. Med Dosim, 30(3), 162{8.
Osei, E. K., Jiang, R., Barnett, R., Fleming, K., & Panjwani, D. 2009. Evaluation
of daily online set-up errors and organ displacement uncertainty during conformal
radiation treatment of the prostate. British Journal Of Radiology, 82(973), 49{61.
Ost, P., Fonteyne, V., De Neve, W., De Gersem, W., De Wagter, C., Vandecasteele,
K., Duprez, F., & De Meerleer, G. 2009. Volumetric modulated arc therapy for
delivery of prostate radiotherapy: In regard to Palma et al. (Int J Radiat Oncol
Biol Phys 2008;70:996-1001). International Journal of Radiation Oncology Biology
Physics, 73(4), 1286{1286.
References
264
Otto, K. 2008. Volumetric modulated arc therapy: IMRT in a single gantry arc.
Medical Physics, 35(1), 310{317.
Otto, K. 2009. Letter to the Editor on 'Single-Arc IMRT?'. Physics in Medicine and
Biology, 54(8), L37{L41.
Paelinck, L., Reynaert, N., Thierens, H., De Wagter, C., & De Neve, W. 2003. The
value of radiochromic lm dosimetry around air cavities: experimental results and
Monte Carlo simulations. Phys Med Biol, 48(13), 1895{905.
Paelinck, L., Reynaert, N., Thierens, H., De Neve, W., & De Wagter, C. 2005. Experimental verication of lung dose with radiochromic lm: comparison with Monte
Carlo simulations and commercially available treatment planning systems. Phys
Med Biol, 50(9), 2055{69.
Palma, D., Vollans, E., James, K., Nakano, S., Moiseenko, V., Shaer, R., McKenzie,
M., Morris, J., & Otto, K. 2008a. Volumetric modulated arc therapy for delivery of prostate radiotherapy: Comparison with intensity-modulated radiotherapy
and three-dimensional conformal radiotherapy. International Journal of Radiation
Oncology Biology Physics, 72(4), 996{1001.
Palma, D., Vollans, E., James, K., Nakano, S., Moiseenko, V., Shaer, R., McKenzie,
M., Morris, J., & Otto, K. 2008b. Volumetric Modulated Arc Therapy (VMAT) for
delivery of prostate radiotherapy: Reduction in treatment time and monitor unit
requirements compared to intensity modulated radiotherapy. Pages S312{S312 of:
50th Annual Meeting of the American-Society-for-Therapeutic-Radiology-and Oncology.
Palma, D., Moiseenko, V., Vollans, E., McKenzie, M., Morris, J., & Otto, K. 2009.
Volumetric modulated arc therapy for delivery of prostate radiotherapy: In regard to
References
265
Palma et al. (Int J Radiat Oncol Biol Phys 2008;70:996-1001) Reply. International
Journal of Radiation Oncology Biology Physics, 73(4), 1287{1287.
Patel, R. R., Orton, N., Tome, W. A., Chappell, R., & Ritter, M. A. 2003. Rectal dose
sparing with a balloon catheter and ultrasound localization in conformal radiation
therapy for prostate cancer. Radiother Oncol, 67(3), 285{94.
Peeters, S. T., Lebesque, J. V., Heemsbergen, W. D., van Putten, W. L., Slot, A.,
Dielwart, M. F., & Koper, P. C. 2006. Localized volume eects for late rectal and
anal toxicity after radiotherapy for prostate cancer. Int J Radiat Oncol Biol Phys,
64(4), 1151{61.
Pollack, A., Zagars, G. K., Smith, L. G., Lee, J. J., von Eschenbach, A. C., Antolak,
J. A., Starkschall, G., & Rosen, I. 2000. Preliminary results of a randomized radiotherapy dose-escalation study comparing 70 Gy with 78 Gy for prostate cancer. J
Clin Oncol, 18(23), 3904{11.
Pollack, A., Zagars, G. K., Starkschall, G., Antolak, J. A., Lee, J. J., Huang, E., von
Eschenbach, A. C., Kuban, D. A., & Rosen, I. 2002. Prostate cancer radiation dose
response: Results of the M. D. Anderson phase III randomized trial. International
Journal of Radiation Oncology Biology Physics, 53(5), 1097{1105.
Prise, K. M., & O'Sullivan, J. M. 2009. Radiation-induced bystander signalling in
cancer therapy. Nature Reviews Cancer, 9(5), 351{360.
Prise, K. M., Schettino, G., Folkard, M., & Held, K. D. 2005. New insights on cell
death from radiation exposure. Lancet Oncology, 6(7), 520{528.
Qi, Zhen-Yu, Deng, Xiao-Wu, Huang, Shao-Min, Zhang, Li, He, Zhi-Chun, Li,
X. Allen, Kwan, Ian, Lerch, Michael, Cutajar, Dean, Metcalfe, Peter, & Rosenfeld,
References
266
Anatoly. 2009. In vivo verication of supercial dose for head and neck treatments
using intensity-modulated techniques. Medical Physics, 36(1), 59{70.
Quach, K. Y., Morales, J., Butson, M. J., Rosenfeld, A. B., & Metcalfe, P. E. 2000.
Measurement of radiotherapy x-ray skin dose on a chest wall phantom. Medical
Physics, 27(7), 1676{1680.
Ramsey, C. R., Seibert, R. M., Robison, B., & Mitchell, M. 2007. Helical tomotherapy
supercial dose measurements. Med Phys, 34(8), 3286{93.
Rancati, T., Fiorino, C., Gagliardi, G., Cattaneo, G. M., Sanguineti, G., Borca, V. C.,
Cozzarini, C., Fellin, G., Foppiano, F., Girelli, G., Menegotti, L., Piazzolla, A.,
Vavassori, V., & Valdagni, R. 2004. Fitting late rectal bleeding data using dierent NTCP models: results from an Italian multi-centric study (AIROPROS0101).
Radiother Oncol, 73(1), 21{32.
Rawlinson, J. A., Arlen, D., & Newcombe, D. 1992. Design Of Parallel Plate Ion Chambers For Buildup Measurements In Megavoltage Photon Beams. Medical Physics,
19(3), 641{648.
Raysearch, Laboratories. 2003. Biological Optimization using the Equivalent Uniform
Dose (EUD) in Pinnacle3.
Ritter, M. 2008. Rationale, conduct, and outcome using hypofractionated radiotherapy
in prostate cancer. Semin Radiat Oncol, 18(4), 249{56.
Rogers, D. W. O. 1984. Low-Energy Electron-Transport with EGS.
Nuclear Instru-
ments & Methods in Physics Research Section a-Accelerators Spectrometers Detectors and Associated Equipment, 227(3), 535{548.
Rogers, D. W. O., Faddegon, B. A., Ding, G. X., Ma, C. M., We, J., & Mackie, T. R.
References
267
1995. BEAM - A Monte-Carlo Code to Simulate Radiotherapy Treatment Units.
Medical Physics, 22(5), 503{524.
Rosenfeld, A. B. 2002. MOSFET Dosimetry on Modern Radiation Oncology Modalities. Radiation Protection Dosimetry, 101(1-4), 393{398.
Rosenfeld, A. B. 2009a (June). The Matreshka: Dual Face-to-Face MOSkin Detector.
Personal Communication.
Rosenfeld, A. B., Siegbahn, E. A., Brauer-Krish, E., Holmes-Siedle, A., Lerch, M. L. F.,
Bravin, A., Cornelius, I. M., Takacs, G. J., Painuly, N., Nettelback, H., & Kron, T.
2005. Edge-on face-to-face MOSFET for synchrotron microbeam dosimetry: MC
modeling. Ieee Transactions On Nuclear Science, 52(6), 2562{2569.
Rosenfeld, Anatoly B. 2009b (May). The Filtering Method for MOSkin Angular Response Correction. Personal Communication.
Rozenfeld, A. B. 2007 (June). Radiation Sensor and Dosimeter (MOSFET)".
Safavi-Naeni, Mitra. 2005. MOSFET Radiation Sensors Reader for Clinical Dosimetry
System. Thesis, Engineering Physics.
Sanghani, M. V., Ching, J., Schultz, D., Cormack, R., Loredo, M., McMahon, E.,
Beard, C., & D'Amico, A. V. 2004. Impact on rectal dose from the use of a prostate
immobilization and rectal localization device for patients receiving dose escalated
3D conformal radiation therapy. Urol Oncol, 22(3), 165{8.
Sanguineti, G., Cavey, M. L., Endres, E. J., Franzone, P., Barra, S., Parker, B. C.,
Marcenaro, M., Colman, M., Agostinelli, S., Foppian, F., & Vitale, V. 2006. Does
treatment of the pelvic nodes with IMRT increase late rectal toxicity over conformal
prostate only radiotherapy to 76 Gy? Strahlentherapie Und Onkologie, 182(9), 543{
549.
References
268
Saunders, M. I., Dische, S., Grosch, E. J., Fermont, D. C., Ashford, R. F. U., Maher,
E. J., & Makepeace, A. R. 1991. Experience with CHART. International Journal
of Radiation Oncology Biology Physics, 21(3), 871{878.
Scalchi, P., & Francescon, P. 1998. Calibration of a mosfet detection system for 6-MV
in vivo dosimetry. Int J Radiat Oncol Biol Phys, 40(4), 987{93.
Scalchi, P., Francescon, P., & Rajaguru, P. 2005. Characterization of a new MOSFET
detector conguration for in vivo skin dosimetry. Med Phys, 32(6), 1571{8.
Schwarz, M., Lebesque, J. V., Mijnheer, B. J., & Damen, E. M. 2004. Sensitivity of
treatment plan optimisation for prostate cancer using the equivalent uniform dose
(EUD) with respect to the rectal wall volume parameter. Radiother Oncol, 73(2),
209{18.
Shah, A. P., Langen, K. M., Ruchala, K. J., Cox, A., Kupelian, P. A., & Meeks, S. L.
2008. Patient dose from megavoltage computed tomography imaging. Int J Radiat
Oncol Biol Phys, 70(5), 1579{87.
Shepard, D. M., Earl, M. A., Li, X. A., Naqvi, S., & Yu, C. 2002. Direct aperture
optimization: A turnkey solution for step-and-shoot IMRT. Medical Physics, 29(6),
1007{1018.
Siochi, R. A. C. 1999. Minimizing static intensity modulation delivery time using
an intensity solid paradigm. International Journal of Radiation Oncology Biology
Physics, 43(3), 671{680.
Skala, M., Berry, M., Duchesne, G., Gogna, K., Tai, K. H., Turner, S., Kneebone,
A., Rolfo, A., & Haworth, A. 2004. Australian and New Zealand three-dimensional
conformal radiation therapy consensus guidelines for prostate cancer. Australas
Radiol, 48(4), 493{501.
References
269
Skwarchuk, M. W., Jackson, A., Zelefsky, M. J., Venkatraman, E. S., Cowen, D. M.,
Levegrun, S., Burman, C. M., Fuks, Z., Leibel, S. A., & Ling, C. C. 2000. Late rectal
toxicity after conformal radiotherapy of prostate cancer (I): multivariate analysis and
dose-response. Int J Radiat Oncol Biol Phys, 47(1), 103{13.
Sohn, M., Yan, D., Liang, J., Meldolesi, E., Vargas, C., & Alber, M. 2007. Incidence
of late rectal bleeding in high-dose conformal radiotherapy of prostate cancer using
equivalent uniform dose-based and dose-volume-based normal tissue complication
probability models. Int J Radiat Oncol Biol Phys, 67(4), 1066{73.
Song, J. S., Court, L. E., & Cormack, R. A. 2007. Monte Carlo calculation of rectal
dose when using an intrarectal balloon during prostate radiation therapy. Med
Dosim, 32(3), 151{6.
Sontag, M. R., & Cunningham, J. R. 1978. Equivalent Tissue-Air Ratio Method
For Making Absorbed Dose Calculations In A Heterogeneous Medium. Radiology,
129(3), 787{794.
Soubra, M., Cygler, J., & Mackay, G. 1994. Evaluation of a Dual Bias Dual Metal OxideSilicon Semiconductor Field Transistor Detector as Radiation Dosimeter. Medical
Physics, 21(4), 567{572.
Spirou, S. V., & Chui, C. S. 1994. Generation of Arbitrary Intensity Proles by
Dynamic Jaws or Multileaf Collimators. Medical Physics, 21(7), 1031{1041.
Spirou, S. V., & Chui, C. S. 1998. A gradient inverse planning algorithm with dosevolume constraints. Medical Physics, 25(3), 321{333.
Steel, G. G., McMillan, T. J., & Peacock, J. H. 1989. The 5rs Of Radiobiology.
International Journal Of Radiation Biology, 56(6), 1045{1048.
References
270
Stein, J., Bortfeld, T., Dorschel, B., & Schlegel, W. 1994. Dynamic X-Ray Compensation for Conformal Radiotherapy by means of Mutli-Leaf Collimation. Radiotherapy
and Oncology, 32(2), 163{173.
Stewart, Bernard W., Kleihues, Paul, Organization, World Health, & for Research on
Cancer, International Agency. 2003. World Cancer Report. IARC.
Storey, M. R., Pollack, A., Zagars, G., Smith, L., Antolak, J., & Rosen, I. 2000.
Complications from radiotherapy dose escalation in prostate cancer: preliminary
results of a randomized trial. Int J Radiat Oncol Biol Phys, 48(3), 635{42.
Suchowerska, N., Hoban, P., Butson, M., Davison, A., & Metcalfe, P. 2001. Directional dependence in lm dosimetry: radiographic and radiochromic lm. Physics
In Medicine And Biology, 46(5), 1391{1397.
Svensson, R., Kallman, P., & Brahme, A. 1994. An Analytical Solution for the Dynamic Control of Multileaf Collimators. Physics in Medicine and Biology, 39(1),
37{61.
Tangboonduangjit, P., Metcalfe, P., Butson, M., Quach, K. Y., & Rosenfeld, A.
2004. Matchline dosimetry in step and shoot IMRT elds: a lm study. Physics in
Medicine and Biology, 49(17), N287{N292.
Teh, B. S., Mai, W. Y., Uhl, B. M., Augspurger, M. E., Grant, W. H., 3rd, Lu,
H. H., Woo, S. Y., Carpenter, L. S., Chiu, J. K., & Butler, E. B. 2001. Intensitymodulated radiation therapy (IMRT) for prostate cancer with the use of a rectal
balloon for prostate immobilization: acute toxicity and dose-volume analysis. Int J
Radiat Oncol Biol Phys, 49(3), 705{12.
Teh, B. S., McGary, J. E., Dong, L., Mai, W. Y., Carpenter, L. S., Lu, H. H., Chiu,
J. K., Woo, S. Y., Grant, W. H., & Butler, E. B. 2002. The use of rectal balloon
271
References
during the delivery of intensity modulated radiotherapy (IMRT) for prostate cancer:
more than just a prostate gland immobilization device? Cancer J, 8(6), 476{83.
Teh, B. S., Dong, L., McGary, J. E., Mai, W. Y., Grant, W., 3rd, & Butler, E. B.
2005. Rectal wall sparing by dosimetric eect of rectal balloon used during intensitymodulated radiation therapy (IMRT) for prostate cancer. Med Dosim, 30(1), 25{30.
Thames, H., & Hendry, J. H. 1987.
and Francis.
Fractionation in Radiotherapy.
London: Taylor
Thames, H. D. 1985. An Incomplete-Repair Model for Survival After Fractionated
and Continuous Irradiations. International Journal of Radiation Biology, 47(3),
319{339.
Thieke, C., Bortfeld, T., Niemierko, A., & Nill, S. 2003. From physical dose constraints
to equivalent uniform dose constraints in inverse radiotherapy planning. Med Phys,
30(9), 2332{9.
Thomas, E., Chapet, O., Kessler, M. L., Lawrence, T. S., & Ten Haken, R. K. 2005.
Benet of using biologic parameters (EUD and NTCP) in IMRT optimization for
treatment of intrahepatic tumors. Int J Radiat Oncol Biol Phys, 62(2), 571{8.
Thomson, I. 1987. Dosimeter.
Tome, W. A. 2009. Seminar 6: Modeling Tumor Control Probability (TCP) from
basic variables of cellular sensitivity. In: Bio-mathematical modeling for cancer
treatments. CMRP.
Tubiana, Maurice. 2009. Can we reduce the incidence of second primary malignancies
occurring after radiotherapy? A critical review. Radiotherapy and Oncology, 91(1),
4{15.
References
272
Tucker, S. L., Dong, L., Cheung, R., Johnson, J., Mohan, R., Huang, E. H., Liu,
H. H., Thames, H. D., & Kuban, D. 2004a. Comparison of rectal dose-wall histogram
versus dose-volume histogram for modeling the incidence of late rectal bleeding after
radiotherapy. Int J Radiat Oncol Biol Phys, 60(5), 1589{601.
Tucker, S. L., Cheung, R., Dong, L., Liu, H. H., Thames, H. D., Huang, E. H., Kuban,
D., & Mohan, R. 2004b. Dose-volume response analyses of late rectal bleeding after
radiotherapy for prostate cancer. Int J Radiat Oncol Biol Phys, 59(2), 353{65.
Tucker, S. L., Dong, L., Bosch, W. R., Michalski, J., Winter, K., Lee, A. K., Cheung,
M. R., Kuban, D. A., Cox, J. D., & Mohan, R. 2007. Fit of a generalized Lyman
Normal-Tissue Complication Probability (NTCP) model to grade >= 2 late rectal
toxicity data from patients treated on protocol RTOG 94-06. Pages S8{S9 of: 49th
Annual Meeting of the American-Society-for-Therapeutic-Radiology-and-Oncology.
Ulmer, W, & Harder, D. 1995. A triple gaussian pencil beam model for photon beam
treatment planning. Z Med Phys, 5, 25{30.
Ulmer, W, & Harder, D. 1996. Applications of a triple gaussian pencil beam model
for photon beam treatment planning. Z Med Phys, 6, 68{74.
Vaarkamp, J., Malde, R., Dixit, S., & Hamilton, C. S. 2009. A comparison of conformal
and intensity modulated treatment planning techniques for early prostate cancer.
Journal of Medical Imaging and Radiation Oncology, 53(3), 310{317.
van Lin, E. N., van der Vight, L. P., Witjes, J. A., Huisman, H. J., Leer, J. W., &
Visser, A. G. 2005a. The eect of an endorectal balloon and o-line correction on
the interfraction systematic and random prostate position variations: a comparative
study. Int J Radiat Oncol Biol Phys, 61(1), 278{88.
References
273
van Lin, E. N., Homann, A. L., van Kollenburg, P., Leer, J. W., & Visser, A. G. 2005b.
Rectal wall sparing eect of three dierent endorectal balloons in 3D conformal and
IMRT prostate radiotherapy. Int J Radiat Oncol Biol Phys, 63(2), 565{76.
van Lin, E. N., Kristinsson, J., Philippens, M. E., de Jong, D. J., van der Vight, L. P.,
Kaanders, J. H., Leer, J. W., & Visser, A. G. 2007. Reduced late rectal mucosal
changes after prostate three-dimensional conformal radiotherapy with endorectal
balloon as observed in repeated endoscopy. Int J Radiat Oncol Biol Phys, 67(3),
799{811.
Vanderstraeten, B., Reynaert, N., Paelinck, L., Madani, I., De Wagter, C., Gersem,
W., De Neve, W., & Thierens, H. 2006. Accuracy of patient dose calculation for
lung IMRT: A comparison of Monte Carlo, convolution/superposition, and pencil
beam computations. Medical Physics, 33(9), 3149{3158.
Vavassori, V., Fiorino, C., Rancati, T., Magli, A., Fellin, G., Baccolini, M., Bianchi,
C., Cagna, E., Mauro, F. A., Monti, A. F., Munoz, F., Stasi, M., Franzone, P.,
& Valdagni, R. 2007. Predictors for rectal and intestinal acute toxicities during
prostate cancer high-dose 3D-CRT: results of a prospective multicenter study. Int J
Radiat Oncol Biol Phys, 67(5), 1401{10.
Veldeman, Liv, Madani, Indira, Hulstaert, Frank, De Meerleer, Gert, Mareel, Marc, &
De Neve, Wilfried. 2008. Evidence behind use of intensity-modulated radiotherapy:
a systematic review of comparative clinical studies. The Lancet Oncology, 9(4),
367{375.
Wachter, S., Gerstner, N., Dorner, D., Goldner, G., Colotto, A., Wambersie, A., &
Potter, R. 2002. The inuence of a rectal balloon tube as internal immobilization
device on variations of volumes and dose-volume histograms during treatment course
References
274
of conformal radiotherapy for prostate cancer. Int J Radiat Oncol Biol Phys, 52(1),
91{100.
Wang, C. W., Chong, F. C., Lai, M. K., Pu, Y. S., Wu, J. K., & Cheng, J. C. H. 2007.
Set-up errors due to endorectal balloon positioning in intensity modulated radiation
therapy for prostate cancer. Radiotherapy And Oncology, 84(2), 177{184.
Wang, J. Z., & Li, X. A. 2003. Evaluation of external beam radiotherapy and
brachytherapy for localized prostate cancer using equivalent uniform dose. Medical Physics, 30(1), 34{40.
Webb, S. 1989. Optimization of Conformal Radiotherapy Dose Distributions by Simulated Annealing. Physics in Medicine and Biology, 34(10), 1349{1370.
Webb, S., & McQuaid, D. 2009. Some considerations concerning volume-modulated
arc therapy: a stepping stone towards a general theory. Physics In Medicine and
Biology, 54(14), 4345{4360.
Webb, Steve. 1993. The Physics of Three Dimensional Radiation Therapy. CRC Press.
Williams, M. J., & Metcalfe, P. 2006. Verication of a Rounded Leaf-end MLC Model
used in a Radiotherapy Treatment Planning System. Phys Med Biol, 51, N65{N78.
Williamson, J. F., Butler, W., DeWerd, L. A., Huq, M. S., Ibbott, G. S., Li, Z., Mitch,
M. G., Nath, R., Rivard, M. J., & Todor, D. 2005. Recommendations of the American Association of Physicists in Medicine regarding the impact of implementing
the 2004 task group 43 report on dose specication for Pd-103 and I-125 interstitial
brachytherapy. Medical Physics, 32(5), 1424{1439.
Withers, H. R. 1975. Advances in Radiation Biology. Academic Press. Chap. The
Four Rs of Radiotherapy.
References
275
Withers, H.R. 1985. Biological Basis for Altered Fractionation Schemes. Cancer, 55,
2086{2095.
Wol, D., Stieler, F., Abo-Madyan, Y., Polednik, M., Steil, V., Mai, S., Wenz, F.,
& Lohr, F. 2008. Volumetric intensity modulated arc therapy (VMAT) vs. serial
tomotherapy and segmental (step and shoot) IMRT for treatment of prostate cancer.
Pages S562{S562 of: 50th Annual Meeting of the American-Society-for-TherapeuticRadiology-and Oncology.
Wong, T. P., Metcalfe, P. E., Kron, T., & Emeleus, T. G. 1992. Radiotherapy x-ray
dose distribution beyond air cavities. Australas Phys Eng Sci Med, 15(3), 138{46.
Woo, M. K., & Cunningham, J. R. 1990. The validity of the density scaling method
in primary electron transport for photon and electron beams. Med Phys, 17(2),
187{94.
Wu, Q., Arneld, M., Tong, S., Wu, Y., & Mohan, R. 2000. Dynamic Splitting of
Large Intensity-modulated Fields. Phys Med Biol, 45, 1731{1740.
Wu, Q., Mohan, R., Niemierko, A., & Schmidt-Ullrich, R. 2002. Optimization of
intensity-modulated radiotherapy plans based on the equivalent uniform dose. Int
J Radiat Oncol Biol Phys, 52(1), 224{35.
Wu, Q., Djajaputra, D., Wu, Y., Zhou, J., Liu, H. H., & Mohan, R. 2003. Intensitymodulated radiotherapy optimization with gEUD-guided dose-volume objectives.
Phys Med Biol, 48(3), 279{91.
Wu, Y., Yan, D., Sharpe, M. B., Miller, B., & Wong, J. W. 2001. Implementing
multiple static eld delivery for intensity modulated beams. Medical Physics, 28(11),
2188{2197.
References
276
Xia, P., & Verhey, L. J. 1998. Multileaf collimator leaf sequencing algorithm for
intensity modulated beams with multiple static segments. Medical Physics, 25(8),
1424{1434.
Xiang, H. F., Song, J. S., Chin, D. W., Cormack, R. A., Tishler, R. B., Makrigiorgos,
G. M., Court, L. E., & Chin, L. M. 2007. Build-up and surface dose measurements
on phantoms using micro-MOSFET in 6 and 10 MV x-ray beams and comparisons
with Monte Carlo calculations. Med Phys, 34(4), 1266{73.
Yan, D., Vicini, F., Wong, J., & Martinez, A. 1997. Adaptive radiation therapy.
Physics In Medicine And Biology, 42(1), 123{132.
Yan, D., Ziaja, E., Jaray, D., Wong, J., Brabbins, D., Vicini, F., & Martinez, A.
1998. The use of adaptive radiation therapy to reduce setup error: A prospective
clinical study. International Journal Of Radiation Oncology Biology Physics, 41(3),
715{720.
Young, M. E., & Kornelsen, R. O. 1983. Dose corrections for low-density tissue inhomogeneities and air channels for 10-MV x rays. Med Phys, 10(4), 450{5.
Yu, C. X. 1995. Intensity-Modulated Arc Therapy with Dynamic Multileaf Collimation
- An Alternative to Tomotherapy. Physics in Medicine and Biology, 40(9), 1435{
1449.
Yu, P. K., Butson, M., & Cheung, T. 2006. Does mechanical pressure on radiochromic
lm aect optical absorption and dosimetry? Australas Phys Eng Sci Med, 29(3),
285{7.
Zelefsky, M. J., Leibel, S. A., Gaudin, P. B., Kutcher, G. J., Fleshner, N. E., Venkatramen, E. S., Reuter, V. E., Fair, W. R., Ling, C. C., & Fuks, Z. 1998. Dose escalation
References
277
with three-dimensional conformal radiation therapy aects the outcome in prostate
cancer. Int J Radiat Oncol Biol Phys, 41(3), 491{500.
Zelefsky, M. J., Fuks, Z., Happersett, L., Lee, H. J., Ling, C. C., Burman, C. M., Hunt,
M., Wolfe, T., Venkatraman, E., Jackson, A., Skwarchuk, M., & Leibel, S. A. 2000.
Clinical experience with intensity modulated radiation therapy (IMRT) in prostate
cancer. Radiotherapy and Oncology, 55(3), 241{249.
Zelefsky, M. J., Fuks, Z., Hunt, M., Lee, H. J., Lombardi, D., Ling, C. C., Reuter,
V. E., Venkatraman, E. S., & Leibel, S. A. 2001. High dose radiation delivered
by intensity modulated conformal radiotherapy improves the outcome of localized
prostate cancer. Journal of Urology, 166(3), 876{881.
Zilio, V., Joneja, O., Popowski, Y., Rosenfeld, A., & Chawla, R. 2006. Calibration of
MOSFET detectors for absolute dosimetry with an (IR)-I-192 HDR brachytherapy
source. Pages 175{175 of: 10th Annual Meeting of the Scientic-Association-ofSwiss-Radiation-Oncology.
Zips, Daniel. 2009. Basic Clinical Radiobiology. A Hodder Arnold Publication. Chap.
Tumour Growth and Response to Radiation, pages 79{101.