Clinical Implementation of IMRT
(TU-B2-01)
Optimization and Delivery of IMRT
• Optimization (Inverse Planning)
• Delivery
• Quality Assurance
• Radiation Safety
Chen-Shou Chui, Ph.D.
Department of Medical Physics
Memorial Sloan-Kettering Cancer Center
New York City, New York
Treatment Planning Procedure
Image Transfer
Pre-optimization
Compute
Contouring
compute dose to point-i
from ray-j, for all i, j
Beam Definition
Optimization
Dose Calculation
a mathematical iterative
optimization process
Plan Evaluation
• Examples of commonly used objective functions:
8target: (D-P) 2
8critical organs: (D-Dc)2, (D-Dc ) if D>Dc
8dose volume conditions:
no more than p% of the volume to exceed dose q.
• Reasonable and mathematically convenient.
• No fundamental physical basis.
Generation of leaf sequencing
file
Conventional planning steps
• Could be any other forms such as |(D-P)m|.
additional steps for IMRT
Examples of Objective Functions
Objective Functions
based on Biological Indices
organ at risk
target
w l(D-P l)
Objective Functions
based on Dose, Dose/Volume
2
(D-Dc)
wu (D-Pu)
2
2
• Maximize TCP, minimize NTCPs.
• Used in place of or in conjunction with dose-,
dose/volume-based objective functions.
• At present, clinical data are scarce and models not
well-established yet.
• Not available on commercial system at present.
Pl Pu
Dc
1
Global vs. Local Minimum
Examples of Optimization Methods
(example of local minima)
Objective Function
1
• Deterministic:
Uniform target dose (100% dose):
1 solution, beams 1&2 equally weighted.
8Gradient, Conjugate gradient.
8Maximum likelihood.
target
2
• Stochastic:
crit.
org.
8Simulated annealing.
8Genetic algorithm.
Global Minimum
IMRT vs Conventional Conformal Plans
Is it
important
?
no
There is
no issue.
yes
IMRT should do better (at least no worse),
for there are more degrees of freedom.
Dose Distributions
Dose Volume Histograms
IMRT vs Conventional Conformal Plans
105
Critical organs: no more than 1/2 to exceed
10% dose.
2 solutions, beams 1&2 weighted either (a,b)
or (b,a) ⇒ local minima !!
Dose Uniformity in the Target
• In Principle:
8Stochastic methods can find global minimum.
8Deterministic methods may get trapped in local minima.
• In Practice:
8It is not easily known if current solution is at global or
local minimum, regardless of which methods are used.
8It may not be important if the solution is clinically
satisfactory.
8If solution not satisfactory:
ý for stochastic methods: change random selection
parameters (e.g.annealing schedule),
increase number of iterations.
ý for deterministic methods: try different initial guess.
ý is the new solution better?
target
Critical organs: no more than 1/2 to exceed
50% dose.
1 solution, beams 1&2 equally weighted.
100
IMRT vs Wedged Pair
100
70
IMRT
80
Volume (%)
45°
45°
Wedged Pair
60
40
target
20
rest of volume
0
0
20
40
60
80
100
120
Dose (%)
Wedged pair
IMRT pair
2
Nasopharynx IMRT and Conformal Plans
Nasopharynx Plan Comparison
IMRT Constraints
Beam arrangement
Target
Minimum: 100%
Uniformity: < 20%
Cord max: 40 Gy
Brainstem max: 45 Gy
Prescription
– Gross disease: ≥ 70 Gy
– Micro. Disease ≥ 54 Gy
– Bid after 36 Gy
7700
7400
7000
6500
Conformal Plan
5000
Normal tissues
–
–
–
–
3500
Cord, brainstem
Mandible
Parotids
Temp. Lobes
2000
LATS TO 36 Gy
Plan for 34 Gy
%VOLUME
100
100
PTV
80
60
40
40
C o n f.
20
CORD
1D Intensity Profiles
80
60
before & after smoothing
C o n f.
IMRT
20
IMRT
0
0
0
2000
4000
6000
8000
0
2000
4000
DOSE (cGy)
100
%VOLUME
IMRT
Conformal
6000
8000
DOSE (cGy)
Savitzky-Golay
(least squares fit)
100
MANDIBLE
TEMP. LOBES
80
80
60
60
Conf.
40
40
IMRT
Conf.
20
20
IMRT
0
0
0
2000
4000
6000
8000
0
2000
DOSE (cGy)
4000
6000
DOSE (cGy)
2D Intensity Distribution
8000
before smoothing
after smoothing
The ‘Skin-Flash’ Problem in Inverse Planning
before & after smoothing
Conventional:
Margin added to field
edge to allow for
uncertainties.
IMRT:
PTV
Intensity remains at 0
outside PTV. No skinflash
before smoothing
after smoothing
3
Skin-Flash for Intensity-Modulated Field
Incorporation of Uncertainties in Inverse Planning
Simple, flat extension
• Each point in the region of
interest (e.g. CTV) is split into
multiple points (typically 3) to
account for the range of
uncertainties.
skin
• Corresponding rays are
brought into optimization.
CTV
• Performed in pre-optimization.
Skin-Flash for Intensity-Modulated Field
Incorporation of Uncertainties in Optimization
Beam’s-Eye-View: Lateral Field
Without skin flash
With 2 cm skin flash
skin
Summary - Optimization
Delivery of IMRT
• Objective functions may be based on dose, dose/volume conditions,
or biological indices.
• Optimization methods may be stochastic or deterministic.
• Local minima may exist, but there is no easy way to tell whether a
solution is at a global or local minimum, regardless of which
optimization method is used.
• Optimized intensity distribution should give better dose distribution
than conventional conformal plans.
• Smoothing of intensity profiles may be useful for practical reasons.
• Other clinical considerations, such as “skin flash” problem.
• Compensator: less efficient (fabrication, re-entry
into the room between fields).
• Fixed field with conventional MLC:
8continuous leaf motion,
8step-and-shoot.
• Rotational field:
8conventional MLC,
8NOMOS/MIMIC, Tomotherapy .
4
Delivery of IMRT with a Conventional MLC
Delivery of IMRT with a Conventional MLC
sliding window method
radiation
Continuous
direction
of motion
Right-leaf
Left-leaf
Beam-on-Time
T
beam-on
time
Step-and-shoot
T
d
Left-leaf
d
Left-leaf
c
c
Right-leaf
delivered intensity
delivered intensity
∆t = 1 ≤ 1
∆x v vmax
a
Right-leaf
b
b
X
P
a
X
P
P
Clinac 2100C 15-MV X-rays
Intensity Modulation by DMLC
Delivery of IMRT with
continuous leaf motion
• Each leaf pair forms a window which slides across
• Dose given through the window as function of MU
•••
•••
1
i
N
1+...+ I +… +N
• The final dose distribution is the summation from
all “segments”
Prostate
Generating MLC Segments
step-and-shoot
Step-and-Shoot MLC Segments
Field shapes and relative intensity weights
Intensity Profile
Intensity Grouping
unconstrained
intensity levels
limit delivery to a few
discrete intensity levels
Beam Segments
includes MLC
constraints
• A bitmap of the optimized beam intensity profile is
converted into deliverable MLC fields, taking MLC
positioning constraints into account.
William Beaumont Hospital
William Beaumont Hospital
5
Variation of Output with Field Shape/Size
Backprojection to the Source Plane
Total intensity at P :
Intensity as a function of
position from the leaf end
beam
Primary Source
Scatter Source
Effects of Rounded Leaf-End and Leaf Transmission
Source
plane
T
φ p = ∫ I(xr (t)-p) I(p-xl (t)) dt
t=0
leaf end x(t)
MLC
opening
left leaf
MLC
plane
P
beam-on
time
direct
exposure
1
right leaf
ε
Right-leaf
Left-leaf
I(x)
leaf
penumbra
leaf
transm.
Isocenter
plane
direction
of motion
x
P
P
Iterative process to generate leaf paths
Organ Motion Relative to Leaf Motion
Desired
fluence profile F(x)
Assign Fwork (x) = F(x)
T
Beam-on-Time
Modify working profile
Fwork(x) = Fwork(x) - E(x)
Parallel
Perpendicular
Calculate leaf paths using Fwork(x)
Calculate generated profile Fg(x)
taking all factors into account
Calculate error E(x) = Fg(x) - F(x)
d
Left-leaf
d’
Right-leaf
c’ c
b’ b
Desired intensity
a a’
error
acceptable
?
finish
Effects of Intra-Fraction Organ Motion on Delivered Intensity
X
Y
P
P
Intra-fraction breathing motion
A = ±3mm, t = 4 sec.
Total intensity at P :
T
φ p = ∫ I(xr (t)-p) I(p-xl (t)) dt
t=0
Static :
Intensity as a function of
position from the leaf end
beam
leaf end x(t)
Organ motion :
f : motion function
t: beam-on-time
τ: period
A: amplitude
to : initial phase
Averaged over 30 fraction
I(x)
P is constant.
P = f(t;τ,A) + t o
Single fraction
direct
exposure
1
ε
leaf
transm.
leaf
penumbra
x
Leaf pair # 14
6
Intra-fraction breathing motion
Splitting a Large IMRT Field
A = ±3mm, t = 4 sec.
105
103
50
IMRT Averaged over 30fx
• Due to the design of MLC,
the largest IMRT size that can
be delivered without jaw
repositioning may be limited
(e.g. 14.5 cm on the current
Varian MLC).
• The most straight forward
solution is to split the field at
an intermediate position Xs, but
it is less efficient (Tl-T r) and
prone to field matching
uncertainties.
T
T
Time
IMRT Planned
100
x2
Field width x2-x1 > 14.5 cm
Splitting a Large IMRT Field (cont’d)
L2
T
Left leaf
Ls
Ts
L1
R2
Rs
Right leaf
F(x)
F2(x)
F1(x)
x2
xl
xr
Split at a beam-on-time
0 x 1R 1
T
L2
Left leaf
A
R2
Right leaf
Tb L1
R F1(x)
0 x1 1
x
B
b
F(x)
F2(x)
xa
Rs
R
0 x 1
1
X
R2
Right leaf
F(x)
xs
Split at a position
x2
Summary - Delivery
• Another way is to split the
field at an intermediate path,
with each segment satisfying
the field size limitation. It
provides a more general
solution but is less efficient in
beam-on-time.
Ta
Ls
L1
0 x
1
• The most efficient way is to
split the field at an intermediate
time Ts, if each segment meets
the field size limitation. It is
most efficient (same beam-ontime) and is insensitive to field
matching uncertainties.
Left leaf
Tr
Right leaf
F(x)
Transverse View
L2
Tl
Left leaf
x2
• Intensity-modulated field can be delivered with a conventional
MLC, either in continuous or step-and-shoot mode.
• Need to account for leaf transmission, rounded leaf end, and
head scatter.
• Increased beam-on-time comparing to conventional treatment:
- prostate, 2-3 times.
- head & neck, 3-4 times.
- breast, about the same.
• Need to account for intra-fraction organ motion, if present:
single fraction & multiple fractions.
• Split large field into 2 or more segments, if necessary.
Split with an intermediate path
Gap error → Dose error
Quality Assurance
20.0
Range of gap width
• Machine Performance :
• Patient Dosimetry:
8Record & verify, file check-sums (each fraction)
8Independent MU check, portal image (each patient)
8Log file analysis, chamber measurement,
film dosimetry (periodically or new technique, new site)
15.0
% Dose error
8Film test (semi-weekly)
8Dosimetry test (monthly)
8Drift test (monthly)
Gap error (mm)
10.0
2.0
1.0
5.0
0.5
0.2
0.0
0
1
2
3
4
5
Nominal gap (cm)
7
Film test
1 mm bands
errors introduced
DMLC Output vs Time
0.170
0.168
DMLC / 10x10
←
← - 0.5 mm
←
← - 0.2 mm
←
← + 0.2 mm
0.2 mm
0.166
0.164
0.162
245
←
← + 0.5 mm
0.160
8
9
8
7
8
8
7
r-9
r-9 un-9
g-9
c-9
c-9
p-9
Ma
Se
De
Ma
J
Au
De
445
Date
Independent Monitor-Unit (MU) Calculation
• Required by regulations, traditionally hand-calculation by the user,
1.03
1.03
not part of the treatment planning system.
1.02
1.02
• For IMRT, such hand calculation is difficult. Independent program
1.01
1.01
is needed to perform MU check.
1.00
0.2 mm 1.00
0.99
0.99
0.98
0.98
245
0.97
8
7
8
7
8
98
t-9 v-9 n-9 pr-9 n-98 ug- ct-9
Oc No Ja
A Ju A
O
Date
• Uses the Beam-on-Time (MU) and leaf sequence file (DVA) as
input, calculate dose distributions in a flat phantom.
• Takes into account rounded leaf-ends, extended source, and leaf
445
0-0
0.97
90-0
270-0
8
7
8
8
7
98
t-9 v-9 an-9 pr-9 un-9 ug- ct-98
180-0
Oc No
J
A
J
A
O
90-90
270-90
Date
transmission.
• Agreement between calculation & measurement, generally < 2% in
dose or < 2mm
• Serves as independent check to treatment planning system.
Nasopharynx - Left Lateral
Measured vs Calculated
Sum of 5 IMRT fields for prostate treatment
Sup
1.03
6 MVx
1.02
Ant
Dmeas / Dcalc
Relative output
DMLC Output vs
Gantry / Collimator Angles
Post
1.01
1.00
0.99
0.98
0.97
0.96
Inf
Film
Plan
mean = 0.993, s = 0.008
0.95
0
245
100
445
200
300
400
Patient #
8
Beam Stability: Dose Rate
Step-and-Shoot
1.04
Beam Stability: Dose Rate
SL20 - 18MV
u
For > 3MU, dose rate is within +/-2% (2s).
u
u
1.03
Relative Dose Rate
u
With step-and-shoot delivery, there is the potential for
short irradiation times (MUs).
Dose rate stability influences the treatment precision.
Measure dose per MU versus total MU.
Check short, and long term stability.
u
Initial
2 weeks later
6 months later,
(new monitor chamber)
1.02
1.01
1.00
0.99
1
10
MU
100
William Beaumont Hospital
William Beaumont Hospital
Beam Stability: Flatness, Symmetry
Beam Stability: Flatness, Symmetry
u Stability of flatness and symmetry affects dose rate
for small fields directed off the central axis.
u F o r a n o p e n 2 0 x 2 0 c m 2 field, measure profiles for
irradiation ranging from 1 to 100 MU.
Symmetry
0.10
u Symmetry is less than +/-3% if more than 3MU
delivered.
Flatness
– Sun Nuclear Profiler (46 diodes, 10 profiles/sec).
u Flatness is less than +/-3% if more than 5MU
delivered.
GUN-TARGET DIRECTION
CROSS-PLANE DIRECTION
0.05
0.00
-0.05
0.10
0.05
0.00
-0.05
1
Radiation Safety Considerations
• Whole Body Dose: (for more details, see Followill et al.
IJROBP.38: 667-672, 1997.)
To estimate the total body dose delivered to patients by photons
and neutrons outside the radiation fields when intensity modulated
beams are used for treatment. These estimates are then used to
compute the risk of induction of secondary cancers as a sequella of
the radiation therapy.
• Room Shielding Requirement: (for more details, see TU-D2-5)
To calculate the shielding required as a result of increased beamon-time due to delivery of intensity-modulated beams.
10
100
Total M U s Delivered
William Beaumont Hospital
William Beaumont Hospital
X-ray and Neutron
whole-body dose equivalent (mSv)
per unit calibration dose (cGy)
Radiation
Type
X-ray Beam Energy
6 MV
18 MV
25 MV
x-ray
8.0 x 10-2
6.5 x 10-3
1.0 x 10-2
neutron
0.0
4.6 x 10-2
7.6 x 10-2
9
MU/cGy to deliver conventional and
modulated beam intensity radiotherapy
Conventional
Beam Energy
Total whole-body dose equivalent (mSv)
for delivered dose of 7000 cGy at isocenter
Beam Intensity Modu lated
unwedged
wedged
Varian MLC
Nomos MLC
(MU/cGy)
(MU/cGy)
(MU/cGy)
(MU/cGy)
6 MV
Conventional
MLC M odul ated
6 MV
1.2
2.4
3.4
9.7
18 MV
1.0
1.5
2.8
8.1
25 MV
1.0
1.5
2.8
8.1
Tomotherapy
18 MV
25 MV
no wedges wedges
no wedges wedges
no wedges wedges
67.
190.
543.
326.
911.
2637.
602.
1686.
4876.
134.
-------
488.
-------
903.
-------
Estimated percent likelihood of a fatal secondary
cancer due to a 7000 cGy course of IMRT
6 MV
Conventional
MLC M odul ated
Tomotherapy
18 MV
25 MV
no wedges wedges
no wedges wedges
no wedges wedges
0.3%
0.8%
2.2%
1.3% 2.0%
3.6%` ---10.5% ----
2.4% 3.6%
6.7% ---19.5% ----
0.5%
-------
Conclusion: Careful consideration should
be made of the implications associated with
secondary whole body radiation before
implementation of beam intensity
modulated conformal therapy using x-ray
energies greater than 10 MV.
Shielding
Primary Barrier
• IMRT = relatively inefficient delivery
• New formalism required
• Evaluate
• Standard:
– workload = TD = MU (50-100K/week) = dose equivalent
• DMLC
– workload = TD (not MU) = dose equivalent
– Primary
– Scatter
– Leakage
• Tomotherapy
– workload = TD * N * F = dose equivalent
• Decouple the workload from MUs
Mallincrodt Inst . of Radiology
• N = average number of indexes
• F = fraction of arc subtended by primary beam
Mallincrodt Inst . of Radiology
10
Secondary Barrier
Secondary Barrier
• Leakage
– Standard
• Patient Scatter
• workload = TD = MU (50-100K/week) = dose equivalent
– Standard:
• workload = TD = MU (50-100K/week) = dose equivalent
– DMLC
• workload = TD * E = MU = dose equivalent
– DMLC
– E = efficiency (approx 1.7)
• workload = TD (not MU) = dose equivalent
– Tomotherapy
– Tomotherapy
• workload = TD (not MU) * α = dose equivalent
• workload = TD * N * E = MU = dose equivalent
– N = average number of indexes
– E = efficiency (approx 2.5)
Mallincrodt Inst . of Radiology
Mallincrodt Inst . of Radiology
Conclusions
Neutron Shielding
• Optimized IMRT plans can produce better dose distributions than
– Standard
• workload = TD = MU (50-100K/week)
– DMLC
conventional conformal plans, in terms of both target coverage and
normal organ sparing.
• Delivery of IMRT with a conventional MLC is practical, either in
continuous or in step-and-shoot mode. Delivery with MIMIC is also
practical, but in general requires longer beam-on-time.
• workload = TD * E = MU
– E = efficiency (approx 1.7)
• Comprehensive QA is needed, for machine performance and
– Tomotherapy
patient-specific dosimetry.
• workload = TD * N * E = MU
• Radiation safety issues need to be considered.
• PROCEED WITH CAUTION !
– N = average number of indexes
– E = efficiency (approx 2.5)
Mallincrodt Inst . of Radiology
Acknowledgment
MSKCC
MD Anderson
Spiridon Spirou
David Followill
Jie Yang
Stanford
Margie Hunt
William
Beaumont
Di Yan
John Wang
Art Boyer
Linda Hong
Tom LoSasso
Mallincrodt
Dan Low
Sasa Mutic
11