Act
Act III
II
Planning
Delivery
Image Processing
for
T
Tx
Delivery
x Planning
Jan-Jakob
Marc Kessler
Sonke
Netherlands
The University
Cancer
of Michigan
Institute
Disclosure
• Our department has research contracts with:
• Elekta Oncology Systems
• Philips Radiation Oncology Systems
• Siemens Medical Solutions
• Our department licenses software to:
• Elekta Oncology Systems
Radiotherapy Procedure
Act III:
Align patient on
machine on tattoos and
treat (many days)
Act I:
Tattoo, align and
scan patient
Act II:
Define the target
and design a
treatment plan
Setup Errors
The patient moves from day to day
Organ Motion
Organs move
from day to day
How can we solve this problem ?
1. Use large margins, irradiating
too much healthy tissues
2. Use small margins, and risk
missing the target
3. Or: use image guided radiotherapy
Image Guided Radiotherapy
• Image the tumor + organs-at-risk or their
surrogates just prior or during treatment
• Assess changes in patient position relative to
treatment plan
• Adapt treatment plan (couch shift) to account for
changes, increasing treatment precision
Safety Margins
Verellen et al. Nature Reviews Cancer 2007
Many In-room Imaging Systems
image guidance: not a new idea !
First isocentric Co-60 machine in Netherlands at NKI (1960)
Visualization
Visualization: Image Fusion
Reference Image
Localization Image
Visualization: Sliding Window
Visualization: Overlay
Complementary color overlay clearly shows gross differences and local
differences provided adequate contrast
Visualization: Animation
Animation clearly shows local changes
Find the differences
Colorwash is not so clear
The power of animation
Pop Quiz
What is the purpose of IGRT
1.
54% 2.
13% 3.
31% 4.
2% 5.
0%
Make pretty images
Minimize setup errors
Quantify organ motion
Reduce PTV margins
Sell more expensive treatment machines
Correct Answer: 5)
Seminars in Radiation Oncology Volume 17, Issue 4
2D Image Guidance
Electronic Portal Imaging
Portal Image Quality
Image quality is limited by:
• Projective nature of 2D
images
A-P
• Low DQE of EPIDs in MV
range
• Limited dose used for
imaging (ALARA)
Lat
.
Prostate
Lung
• Limited contrast differences
of tissue in the MV range
Mostly limited to bony anatomy alignment
minimizing setup errors
Preprocessing: unsharp masking
Top-hat transform
Original image
Binary top-hat enhanced image
Portal image analysis - 2D
Reference image Match field-edge Match anatomy
Setup Error = Anatomy Match – Field-Edge Match
3D EPID Dose reconstruction
prostate VMAT plan
EPID movie
• Energy: 10 MV
• 243 frames
• delivery time: 96 s
Dose per frame
Accumulated dose
axial slice through isocentre
Prostate VMAT plan
EPID reconstructed
3D dose distribution
3D -evaluation
EPID vs. plan
axial slices
50% isodose
Disoc EPID: 217.3 cGy
Disoc plan:
219.2 cGy
difference: -0.8%
0.60
1% 0.95
P ≤1 99%
mean
3D Image Guidance
3D
match
Cone beam CT
Planning CT
same ?
MV image
DRR + Template
2 x 2D
match
AP/LAT
Evaluation of Bony Registration
• Tests comparing EPID and CBCT for
checking bony anatomy setup:
• Good results for prostate and head and neck
• Poor results for lung
• Aim: Cross-validation of setup measurement
using CBCT and EPID for lung.
• 8 patients, 66 fractions total
Matching cone beam to planning CT on bone is highly accurate example for lung treatment series - 10 days matched
Vertebrae are
perfectly still
Estimated match accuracy << 1 mm SD, much better as EPID for lung
Comparison of cone-beam CT (purple)
with EPID (green) for 10 fractions
No patient motion visible on spines; some respiration level differences
Correlation cone-beam CT - EPID
Ant-Post
10.0
10.0
10.0
8.0
8.0
8.0
6.0
6.0
6.0
4.0
4.0
4.0
2.0
0.0
-5.0
0.0
5.0
-2.0
10.0-10.0
2.0
0.0
-5.0
0.0
-2.0
5.0
EPID
-10.0
Sup-Inf
EPID
EPID
Left-Right
10.0
-10.0
2.0
0.0
-5.0
0.0
5.0
-2.0
-4.0
-4.0
-4.0
-6.0
-6.0
-6.0
-8.0
-8.0
-8.0
-10.0
-10.0
-10.0
Setup error measured with cone beam CT
Slope: 0.81 .. 1.06
0.66 .. 0.88
0.30 .. 0.66 (95% CI)
There is a clear under-estimation of setup error with the EPID !
10.0
Μαρχ Κεσσλερ / ΥΜ
How many Degrees of Freedom?
ΠΕΤ/ΧΤ
ΜΡ − ΧΤ
4Δ ΧΤ
0?
3 to 6
3xN
Νονε ?
Φεω
Μανψ
Requirements for IGRT
registration
• Fast and robust image registration
• Easy visual validation
• Registration result drives a couch shift
Rigid registration with 6 degrees of freedom is
a likely candidate
Automatic matching on region of
interest built-in in Synergy system
reference
localization
Tumor in top of neck
Required table shift:
(-3.2, -1.5, -0.6) mm
reference
localization
Tumor in lower part of neck
Required table shift:
(+1.5, -3.2, -6.1) mm
Rotations
Rotations
Correction by Couch Shift
Modify Rotation Point
Correction by Couch Shift
Soft Tissue Guidance
Grey-value registration
TAP / TCC / TLR / RAP / RCC / RLR
** Smitsmans et al.,
IJROBP 60 (2004)
Automatic prostate localization in CBCT (30 s)
Cone beam CT
10 CBCT scans: automatic bone match
Planning CT contours
placed automatically
10 CBCT scans: automatic prostate match
help line (GTV+3.6 mm)
Smitsmans et al., IJROBP 2004, 2005
Pop Quiz
How many degrees of freedom are typically
used for IGRT image registration
1.
0% 2.
32% 3.
59% 4.
2% 5.
0
3
6
42
Not enough
7%
Correct Answer: 3)
Van Herk et al. Seminars in Radiation Oncology, 2007
4D Image Guidance
The CT imaging problem
Spirometry – Thermo-couple
• Thermo-couple measures
warm & cold air flow
• In its simple form relative signal
no quantative information
1
200
0.9
150
R
espirationsignal -R
aw
0.7
50
0.6
0
0.5
0.4
-50
0.3
-100
0.2
-150
0.1
-200
0
10
20
30
40
50
60
70
80
0
R
espirationsignal -Scaled
0.8
100
Sort slices
(1) Reconstructing many slices (2) Sorting CT slices
Selected slices gathered,
yielding a single phase CT
Selection
Raw CT
Raw CT with
respiration signal
4D Computer Tomograpy
Breathing
Respiratory signal extraction
Vertical
derivative filter
Horizontal
projection
Temporal
concatenation
[Zijp, ICCR, 2004]
[van Herk, ICCR, 2007]
Amsterdam shroud (2D image)
RCCBCT
3D versus 4D CBCT
• 4D Data set
• 8 x 84
projections
• 3D Data set
• 670 projections
ROI by GTV Expansion
4D CBCT + GTV Contour
Local Rigid Body Registration
Visual Validation
Apply Correction
Impact of Respiratory Motion
on Dose Distribution
Planned distribution
Delivered distribution
• Shift of the dose distribution due to
displacement of the mean tumor position
• Blurring of the dose distribution due to
breathing around the mean position
Pop Quiz
Which motion management strategy has the
largest impact on the delivered dose:
1.
16%
2.
7%
3.
24%
4.
50%
5.
4D Imaging
4D Planning
4D Delivery
4D Image Guidance
The impact of respiration is over exaggerated
4%
Correct Answer: 4)
Guckenberger et al. Radiother. Oncol. 2009
Adaptive Radiotherapy
Differential Variability
No couch correction can solve this problem
Repeat 4D cone beam CT
Shows respiration, tumor shrinkage and baseline position variation
Anatomical Changes
Bony anatomy
registration
non-rigid registration
Fraction at day # 36
Results, volume change of target + OAR
day
0
1
7
14
21
28
36
CTV
387
381
378
373
373
344
315
PTV
1051
1036
1035
1008
1012
969
916
Spinal cord
20
20
20
20
20
21
20
Parotid Li
31
31
32
27
28
23
18
Parotid Re
21
21
21
20
19
17
15
t(s)
67
120
104
55
141
155
Deformable Registration
Summary
• Geometrical uncertainties limit the precision of
radiotherapy
• 2D,3D&4D image registration and guidance
increases the precision of RT allowing margin
reduction and dose escalation
• Adaptive RT further individualized treatment
delivery
Acknowledgements
Marcel van Herk
Peter Remeijer
Jochem Wolthaus
Simon van Kranen
Simon Rit
Monique Smitsmans
Anton Mans
David Jaffray
Doug Mosely
Mark Kessler