PATIENT/ORGAN GEOMETRY VERIFICATION
IN EXTERNAL BEAM TREATMENT DELIVERY
Di Yan, D.Sc
Department of Radiation Oncology
William Beaumont Hospital, Royal Oak, Michigan
One of major impediments to ensure the accuracy of radiation delivery and to
achieve treatment optimization is the inherent variability of patient and organ geometry
over the course of radiation dose delivery. To overcome this obstacle, advanced imaging
technologies to facilitate patient/organ geometry detection during treatment delivery are
being rapidly developed. These technologies include MV portal imaging for detecting
daily patient setup error, and on board ultrasound and CT imaging for detecting internal
soft tissue motion. In addition, embedded radiomarker displacement monitored via online kV radiographic imaging, as well as body surface motion monitored using a CCD
camera, have been investigated as surrogates for measuring inter or intra-treatment organ
motion respectively.
In tandem advanced imaging technology, computer algorithms and software tools
for measuring patient/organ geometric variation relative to the radiation beam have been
developed. These algorithms can be divided into two groups: Group I for determining the
rigid body motion of patient bony structure or inserted radiomarkers, and Group II mainly
for determining the non-rigid body motion of patient soft tissues. Temporal variation of
patient/organ geometry during radiation treatment has been classified, based on its
statistic behavior, as a systematic variation and a random variation. The effects of these
two variations on treatment planning are quite different, and the systematic variation is
dominant factor. One should carefully examine the effects in relation to the treatment goal
and clinical resources before selecting a method to correct and compensate for them.
The major goal of measuring patient/organ geometric variation can be twofold.
The first is for treatment quality assurance, where patient/organ geometric variation
identified from image feedback is corrected to ensure that the treatment can be delivered
as pre-planned. Various methods of off-line and on-line patient/organ position correction
have been proposed and some of them have been clinically implemented. Additional
study of output residuals is critical and required to validate these correction methods. The
second purpose of using image feedback is more proactive, whereby the feedback images
are systematically applied in designing an optimal dose distribution so that treatment
planning and decisions can be fully adapted to patient/organ geometric variation. Few
strategies have been proposed, so far, and only for limited treatment sites. The major
challenges in implementing an adaptive control strategy include image-based organ
registration, organ subvolume position estimation, and beam intensity optimization. In
addition, practical issue has heavily influenced the new developments on this subject.
S
l
i
d
e
Patient/Organ Geometric Variation
In Radiotherapy Process
physical patient
2
patient
1x
CT room
Lasers
■ Skin markers
■ Images
■ Bone
■ Tumor/organ
■ Delineation
■ Margin
■ Planned beam
Treatment room
Lasers
■ Skin markers
■ Bone
■ Tumor/organ
■
■
■
■
Beam
Accelerator
■ Treatment room
Nx
■
■
Nx
beam data
Courtesy Marcel Van Herk
S
l
i
d
e
Systematic And Random
Displacements
Simulation
Treatment Delivery
x0
v
x t : t = 1 , ⋅ ⋅ ⋅, n
v
v
3
v
µ ( xv 0 , xvt :t =1,⋅⋅⋅, n ) ; σ
v
( x t : t = 1,⋅ ⋅ ⋅, n )
There have been numerous steps
in a radiotherapy process, which
potentially cause geometric
discrepancy on patient/organ
position between treatment
planning and actual treatment
delivery. In treatment simulation
& planning, starting from patient
positioning and ending at
transferring treatment plan to
radiotherapy machine, each step
will introduce uncertainties in
patient/organ geometry. On the
other hand, most of the
positioning steps will be repeated
during the treatment delivery
introducing the systematic
variation and random variation in
the patient/organ treatment
geometry.
Patient/organ geometric variation
can be best quantified using
subvolume displacement in an
organ of interest. Subvolume of
an organ of interest has its single
position in treatment simulation,
and multiple positions in actual
treatment deliveries. The
difference between the mean of
the treatment positions and the
simulation position has been
defined as the systematic
displacement of the subvolume,
meanwhile the standard deviation
of the treatment positions has
been used to represent the random
displacement.
S
l
i
d
e
Organ geometric variation can be
quite heterogeneous. The color
map on this figure shows the
distribution of random
displacement represented using
the standard deviation along the
AP direction for a typical patient
prostate and seminal v. Therefore,
eventually we should expect a
non-uniform CTV-to-PTV
margin.
Distribution of Random Displacement in AP (cm)
4
* Patient/organ geometric variation can be a local issue &
heterogeneously distributed
S
l
i
d
e
Effect On Treatment Dose
100
Lung Side
Mediastinal Side
90
80
5
70
Dose Profile
60
50
40
Dose(%)
30
Dose Gradient (%per mm)
20
Dose Curvature (%per mm^2)
10
0
-100
-80
-60
-40
-20
0
20
40
60
80
100
-10
Position (mm)
8
Dose(%)
Dose Grad ient (% per m m )
6
v
µ
Dose Profile
S
l
i
d
e
Dose Curvature (% p er m m ^2)
4
v
σ
2
-100
v
0
σ
v
µ
-2
6
-4
-6
-8
Po sitio n (m m )
v
v
v
∆D = µ T ⋅ ∇D[ xR , µ + xR ]
+
v
σT ⋅
∂2 D v v
v ( xR ) ⋅ σ
∂x 2
* Dosimetric effect of patient geometric variation (with respect
to the corresponding systematic and random displacements) is
mainly dependent on the Dose Gradient & Dose Curvature
100
Dose deviation caused by organ
geometric displacement is also
dependent on the dose distribution
itself. The figure shows a curve of
dose profile (along the patient RL
direction) in a lung cancer
treatment. In this case, dose
profile closed to mediastinal
region has sharper gradient
compared to the lung side due to
the electron transport. Depending
also on the dose prescription
point, the dose gradient (pink) and
dose curvature (blue) around the
target edges will be different.
Dose deviation caused by
subvolume displacement is
mainly dependent on the dose
gradient with respect to the
systematic displacement, and the
dose curvature with respect to the
random displacement. Therefore,
in this specific treatment, we can
expect a larger margin required at
target edge in the mediastinal side
than the one in the lung side.
7
This figure shows the
corresponding margins required
to compensating for the random
displacement alone for both lung
and mediastinal sides of target
edge. The margin is calculated
based on the dosimetric tolerance
of 0.1%. The results indicate that
CTV-to-PTV margins can be
quite different even patient/organ
geometric variations are identical.
CTV-to-PTV margins are
treatment site and treatment
modality dependent.
Effect On CTV-to-PTV Margin
CTV-to-PTV Margin (mm)
S
l
i
d
e
11
10
9
8
7
6
5
4
3
2
1
0
For Random at Lung Side
For Random at Mediastinal Side
2.5
5
7.5
10
12.5
15
sd (σ) of Random Displacement (mm)
* CTV-to-PTV margin is patient and treatment
site specific
8
7
σ
CTV-to-PTV Margin (mm)
S
l
i
d
e
For Gaussian Distribution
6
For Breathing Distribution
5
4
3
2
1
0
2.5
5
7.5
σof RandomDisplacement (mm)
The systematic (µ) and random (σ) displacements are
appropriate representations for patient/organ geometric
variation
10
The effect of patient/organ
geometric variation on physical
dose deviation can be well
determined using the systematic
and random displacements alone.
The figure on left shows two
displacement distributions. The
top one is a Gaussian type, and
the second one is formed from a
respiratory motion. The two
distributions have quite different
shape, however as long as they
have the same systematic
displacement and same standard
deviation of random
displacement, the corresponding
CTV-to-PTV margin required
(left figure) will be very similar.
S
l
i
d
e
9
Summary
• Effect of patient/organ geometric variation on
treatment dose in an organ subvolume can be
appropriately studied using 4 parameters,
∂2D
( µ , σ ) ; (∇ D , v 2 )
∂x
v v
s
• CTV-to-PTV margin is dominated by the
systematic displacement
• It is also treatment site and treatment modality
specific
S
l
i
d
e
Detection Methodologies
Interna l O rga n M otio n
M e th od
F ilm
1
0
MVI
KVI
P atie nt S etup
A natom y
M a rk er
A natom y
M a rk er
A natom y
M a rk er
G ating w ith
C C D & S kin
M a rk er
U ltra so und
Im a ging
M VCT
KVCT
UA , LE
LA, LE
UA, M E
LA, M E
UL, H E
LA , HE
X
R igid
N on R ig id
X
X
X
X
X
X
X
X
X
LA, M E
LA, M E
UA , LE
UA, HE
* UA - Universally Applicable; LA - Limited Applicability
* LE, ME, & HE - Low, Moderate, & High Efficacy
S
l
i
d
e
1
1
Correction Methodologies
Method
Move Couch
Move Couch, Collimator,
Gantry
Adjust Beam
Aperture
Static
Dynamic
Adjust Beam Intensity
fluence
Translational
Rotational
Non Rigid
Motion
Motion
Motion
á
á
X
X
á
X
á
á
á
á
á
á
á
X
á
S
l
i
d
e
1
2
S
l
i
d
e
1
3
Correction Strategies
• Offline
– Reduce CTV-to-PTV margin by reducing the
systematic displacement and customize individual
margin
• Limited measurements
• Estimate the systematic and random displacements
• Online
– Reduce CTV-to-PTV margin by reducing daily
displacements
• Perform measurement and correction before
treatment delivery
Residuals
• Residual (R1) caused by the calibration between an
imaging device and dose delivery device
– Ultrasound
• WA Tome, et al, (UWM), Medical Physics, 2002
• LG Bouchet, et al, (UFG), PMB, 2001
– KVI
• JM Balter, et al, (UM) IJROB 1995
• H Shirato, et al, IJROB, 2000
• L Pisani, et al, (WBH), IJROB 2000
– Gating Device
• G Mageras, et al (MSKCC)
– KV (CB) CT
• C. Rowbottom, et al, (WBH), IJROB 2002
• R1 < 1 mm
S
l
i
d
e
1
4
Residuals
• Residual (R2) introduced by an indirect
measurement
– Marker position vs ROI ?
– Skin motion (gating) vs tumor position ?
• Need more validation study
– Multiple CT scans with markers
– Inserted markers in tumor with skin
motion
S
l
i
d
e
1
5
Residuals
• Residual (R3) introduced by a measurement method
– Center of mass displacement
• Effect of rotation and deformation
– M Van Herk, et la: IJROBP, 33:5,1995
– ~ 10 degree (95% confidence) => 4~5 mm
residual in AP direction at prostate apex and base
– Rigid body transformation (2D)
• Effects of out-of-plane rotations
– J Hanley, et la: IJROBP, 33:5,1995
– P Remeijer, et la: IJROBP, 46:5, 2000
– ~ 4 mm residual (95% confidence)
S
l
i
d
e
1
6
Residuals
• Residual (R3) introduced by a measurement
method
– Rigid body transformation (3D)
• Effects of organ deformation?
– Ultrasound Imaging
• Effect of different image modality (subjective)
– J Lattanzi, et la: IJROBP, 43:4,1999
– 3+1.8 mm (A/P), 2.4+1.8 mm (LAT), 4.6+2.8 mm (S/L)
S
l
i
d
e
1
7
Residuals
• Residual (R4) caused by an estimation
(offline)
– Estimate the systematic and random
displacements
• The residual is dependent on the number
of measurements & random feature
S
l
i
d
e
1
8
S
l
i
d
e
1
9
Residuals
• Residual (R5) due to the limit of a correction
method
– Adjust couch position alone ?
– Adjust couch and collimator alone ?
• Residual (R6) due to the inaccuracy of a dose
delivery device
– MLC position
– Dose output
Summary
•Patient/organ geometric verification by means of image
feedback has two major goals, (a) treatment QA and (b)
treatment optimization by adapting treatment plan to the
feedback information
•No matter which measurement and correction methods
are applied, we should always perform an intensive study
to determine the potential residuals for proper margin
design
S
l
i
d
e
2
0
Acknowledgement
My colleagues in William Beaumont Hospital
Marcel van Herk & Joos Lebesque in Avl/NKI
Gig Mageras & Michael Lovelock in MSKCC
James Balter & Dale Litzenberg in U Michigan
Gustavo Olivera & Wolfquang Tome in UWM
Vincent Khoo in Christie Hospital, UK