Improving Radiation Therapy for Moving TumorsImproving Radiation Therapy for Moving Tumors Outline Brief description of current radiotherapy successes and challenges Dual-Gating: Increasing

Improving Radiation Therapy for Moving Tumors

Sarah Geneser


Outline

Brief description of current radiotherapy successes and challenges

Dual-Gating: Increasing delivery speed without sacrificing targeting accuracy-- Demonstration on a lung cancer case-- Demonstration on a phantom with varying degrees of motion

Examining the effect of deformable image registration on dose warping

Concluding remarks


Radiation Therapy

• destroy tumor cells using radiative energy

• goal: irradiate the tumor (improves tumor control) while limiting dose to surrounding healthy tissues (reduces side effects)

• new hardware makes it possible to shape the beam aperture and deliver high levels of dose to the tumor while keeping normal tissue dose low (highly-conformal therapy)

multi-leaf collimator

(Image courtesy of Varian)

fluence mapconformal dose


Motion Presents Challenges

• several types of motion exist that complicate radiation therapy:• intrafraction:

• respiratory motion - diaphragm displaces abdominal and thoracic organs• prostate motion - prostate can move during treatment (due to bladder and

rectal filling)• interfraction:

• bladder filling - difficult to ensure consistent bladder fill level for treatment• weight loss - can occur over the course of several treatments

• cannot accurately predict intrafraction and interfraction motion, so must develop methods to account for these types of motion• e.g. respiratory motion uses “motion envelope” or gated therapy


Treatment in the Presence of Respiratory Motion

• motion envelope• benefits: ensures tumor coverage, delivery throughout breathing cycle• drawbacks: increases radiation dose to surrounding tissues

• respiration-gating• benefits: reduces irradiation of dose to surrounding tissues• drawbacks: tumor coverage somewhat less certain, delivery during only a

portion of breathing cycle

beam enabled

beam held

exhale window!

Both methods turn the 4D problem into a 3D

problem


4D Therapy in the Presence of Respiratory Motion

• create individual dose plans for each respiratory phase by optimizing over all phases simultaneously

• benefits: ensures tumor coverage while limiting radiation dose to surrounding tissues, delivery throughout breathing cycle

• drawbacks: increases delivery complexity

4DRT work began in 2004, but has not entered the clinic because not

feasible with current linac hardwareRelated Publications:Keall, et. al, Phys. Med. Biol., 49(16), 2004. Rietzel, et. al, IJROBP, 61(5), 2005.Trofimov, et. al, Phys. Med. Biol., 50, 2005.Lee, et. al, Phys. Med. Biol., 54, 2008.


Can We Do Better? Delivering at Inhale and Exhale

• compromise between 4DRT and conventional gated therapy

• benefits: • ensures tumor coverage while limiting radiation dose to surrounding

tissues, • faster than conventional gated therapy• less complicated than than 4DRT delivery

• drawbacks: • more complicated than conventional gated therapy• slower than 4DRT delivery


Dual-Gated Alternating Delivery

inhale window

beam enabled

beam held

exhale window

exhale fluence inhale fluencealternate delivery of IMRT fields


Machine Dynamics: Implementing Dual-Gating

True-Beam XML: used for developing

novel delivery methods on the True-Beam linac

Enables control of the gating windows:


minimize

w

X

s

r

spM

s

kAs

w �D

s

k22

subject to 0 w w

max

Dmin

s

A

s

w Dmax

s

Conventional Treatment Planning Optimization

How do we model dose for dual-gated therapy?

minimum and maximum dose constraints

non-negative fluence constraint

prescribed dose

calculated dose

sum over structures of interest

relative importance weighting normalized by voxels in a structure

IMRT fluence map


Dose Warping and Accumulation

exhale fluence inhale fluence

inhale CTexhale CT

exhale dose

inhale dose

R

registeredinhale dose

R

+

summed dose

=

Dtotal

= R(Dinhale

) +Dexhale


Machine Dynamics

inhale window

beam enabled

beam held

exhale window

exhale fluence inhale fluenceMLCs must move to correct positions between phases


Leaf Motion Regularization

exhale fluence inhale fluence

�

NfX

f=1

NuX

u=2

NvX

v=2

(|wu,v,f,i � wu,v,f,e|) penalizes large pixel-wise differences between the inhale and exale IMRT maps


Dual-Gated Treatment Planning Optimizationdose matching term

leaf motion penalization term TVR term

non-negative fluence constraintminimum and maximum dose constraints

minimize

w1,2

X

s

�

s

kDinhale

+D

exhale

�D

prescribed

k22

+

0

@NfX

f=1

NuX

u=2

NvX

v=2

�(|wu,v,f,1 � w

u,v,f,2|) +2X

p=1

�(|wu,v,f,p

� w

u�1,v,f,p|+ |wu,v,f,p

� w

u,v�1,f,p|)

1

A

subject to 0 w w

max

Dmin

s

ˆ

A1,2,sw1,2 Dmax

s


Evaluate Dual-Gating on Lung Patient Case


Evaluating Dual-Gating Performance (1, 2, 3 cm Motion)

gafchromic film can be placed in the cedar cylinder perpendicular to motion

respiratory-motion stage translates the phanom in the SI direction

right lungheart

left lung

spinalcord

ptv


1 cm SI Translation Dose Plan

+

inhale dose exhale dose

total dose

=

Gy Gy

Gy

0246810

15

21

27

33

39

0246810

15

21

27

33

39

05101520

30

40

50

60



+


Gy Gy

Gy

0246810

15

21

27

33

39

0246810

15

21

27

33

39

05101520

30

40

50

60

total dose

=



+


Gy Gy

Gy

0246810

15

21

27

33

39

0246810

15

21

27

33

39

05101520

30

40

50

60

total dose

=


−1

−0.4

−0.1

0.1

0.4

−1

−0.4

−0.1

0.1

0.6

−1.4

−0.6

−0.10.1

0.8

Dose Differences

1 cm translation 2 cm translation

3 cm translation stationary dosedual-gated plan has greater dose than conventional plan

dual-gated plan has less dose than static plan

05101520

30

40

50

60


Dual-Gating: Conclusions

• produces conformal dose distributions for lung cancer patient and for phantom with up to 3cm motion

• provides 1.75 to 2.23 times speedup depending on patient breathing

• promising option for treating respiratory-gated patients


Dual-Gating: Limitations (and Potential Solutions)

• MLC maximum speed and patient breathing characteristics may complicate delivery and reduce efficiency gains -- possible to encourage brief pauses at inhale and exhale using respiratory coaching

•the treatment for inhale or exhale may finish sooner -- possible to incorporate patient breathing dynamics into optimization to weight one phase more heavily


But ... What About Image Registration?

• Image registration deformations are used to warp dose.

• What if deformations are inaccurate?

• How do these errors effect registered dose?

• How do we determine ground truth?


Neil Kirby’s 2D Deformable Phantom

deformable phantom with optical markers to measure deformations


Prostate Dose Warping and Accumulation

empty bladderfull bladder

What are the errors in the dose warping?

• HDR planned on full bladder and IMRT planned on empty bladder

• Would be useful to be able to sum the two plans to see total dose

1) register full bladder image to empty bladder2) apply transformation to full bladder dose

full bladder registered to empty bladder

R

0

5

10

15

20

25

30

35

40

45

50

55

Gy


Percent Dose Deviations using DIR

symmetric force Demons

−30

−20

−10

0

10

20

30%

%

%

%

%

%

%

MIMVista

−30

−20

−10

0

10

20

30%

%

%

%

%

%

%original Lucas Kanade

−30

−20

−10

0

10

20

30%

%

%

%

%

%

%

free form by Lu

−30

−20

−10

0

10

20

30%

%

%

%

%

%

%


Dose Underestimation and Overestimation

0 5 10 15

10−4

10−3

10−2

10−1

100

Underdosing [Gy]

Frac

tion

of v

oxel

s w

ith g

reat

er u

nder

dosi

ng

lucasKanadeOriginalhornAndSchunckOriginalhornAndSchunckInverseConsistencyiterativeOpticalFlowdemonsFastiterativeOpticalFlowFastdemonsFastWithElasticRegularizationfreeFormDeformationByLudemonsSymmetricForcelucasKanadeImprovedMIMVista

0 5 10 15

10−4

10−3

10−2

10−1

100

Overdosing [Gy]

Frac

tion

of v

oxel

s w

ith g

reat

er o

verd

osin

g

lucasKanadeOriginalhornAndSchunckOriginalhornAndSchunckInverseConsistencyiterativeOpticalFlowdemonsFastiterativeOpticalFlowFastdemonsFastWithElasticRegularizationfreeFormDeformationByLudemonsSymmetricForcelucasKanadeImprovedMIMVista

MIMVista


Lessons Learned

• even when image registration is visually excellent, deformations may still be inaccurate (you can’t necessarily trust the clinical software!)

• difficult for image registration tools to infer deformations in regions of homogeneous dose

• regions of high dose gradient perpendicular to deformation vectors produce large errors in dose


Next Steps?

• Calculate and determine the effect on the dose volume histogramss

• Examine 3D deformations

• Incorporate tissue mechanics into image registration methods. (e.g. Kristi Brock et. al. -- MORFEUS)


Acknowledgements• Dual-Gated Delivery: Lei Xing, Benjamin Fahiminan, Kayla Keilar

• Dose Warping Accuracy: Neil Kirby, Jean Pouliot

• Funding: National Cancer Institute (T32 CA09695)


Thanks! Questions?

inhale window

beam enabled

beam held

exhale window

0

5

10

15

20

25

30

35

40

45

50

55

+ = ??

−30

−20

−10

0

10

20

30


EXTRA SLIDES


Dose Accumulation

Ai,e,swi,e = [Ai,s Ae,s]

wi

we

�= [Ai,swi +Ae,swe]

= di + de

voxel-wise dose at inhale voxel-wise dose at exhale


Dose Operators

voxel-wise dose at inhale voxel-wise dose at exhale

d = di + de

to sum properly, doses must be voxel-wise consistent dose operators must also be voxel-wise consistent)Ii Ie

~xi

T (~xi)

T (Ii)

~xe

1) register images to obtain voxel-wise correspondence

original inhale system: original exhale system:reordered inhale system:

2

66664

3

77775

2

66664

3

77775

2

664

3

775 =

Ai widi 2

66664

3

77775

2

66664

3

77775

2

664

3

775 =

Ai widi 2

66664

3

77775

2

66664

3

77775

2

664

3

775 =

Ae wede

inhale dose

at voxel ~xe

inhale dose

at voxel ~xi

exhale dose

at voxel ~xi

2) reorder dose operators according to the deformation mapping


Dual-Gated Treatment Planning Optimization

minimize

wi,e

X

s

�

s

k ˆ

A

i,e,s

w

i,e

�D

s

k

subject to 0 w w

max

Dmin

s

ˆ

A

i,e,s

w

i,e

Dmax

s

inhale fluence

exhale fluence

wi,e =

wi

we

�Ai,e,s = [Ai,s Ae,s]

inhale dose operator exhale dose operator


−15

−10

−5

0

5

10

15

symmetric force DemonsGy

Warped Isodose Contours using DIR

−15

−10

−5

0

5

10

15

MIMVistaGy

−15

−10

−5

0

5

10

15

original Lucas KanadeGy

−15

−10

−5

0

5

10

15

free form by LuGy

Documents

Improving Radiation Therapy for Moving TumorsImproving Radiation Therapy for Moving Tumors Outline Brief description of current radiotherapy successes and challenges Dual-Gating: Increasing