intensity(modulated/radiotherapy/ - harvard university · 2016. 7. 13. · overview/...
TRANSCRIPT
![Page 1: Intensity(modulated/radiotherapy/ - Harvard University · 2016. 7. 13. · Overview/ Fluence!map!op.mizaon! Extensions!in!IMRT!planning! < Intensity](https://reader033.vdocument.in/reader033/viewer/2022052104/603ee721b6e171642b292bd8/html5/thumbnails/1.jpg)
ì Intensity-‐modulated radiotherapy Beyond fluence map op.miza.on
![Page 2: Intensity(modulated/radiotherapy/ - Harvard University · 2016. 7. 13. · Overview/ Fluence!map!op.mizaon! Extensions!in!IMRT!planning! < Intensity](https://reader033.vdocument.in/reader033/viewer/2022052104/603ee721b6e171642b292bd8/html5/thumbnails/2.jpg)
Overview
Fluence map op.miza.on
Extensions in IMRT planning
-‐ Intensity-‐modulated proton therapy (IMRT)
-‐ leaf sequencing -‐ direct aperture op.miza.on -‐ VMAT op.miza.on
So far:
Now:
![Page 3: Intensity(modulated/radiotherapy/ - Harvard University · 2016. 7. 13. · Overview/ Fluence!map!op.mizaon! Extensions!in!IMRT!planning! < Intensity](https://reader033.vdocument.in/reader033/viewer/2022052104/603ee721b6e171642b292bd8/html5/thumbnails/3.jpg)
Fluence map
IMRT / photons IMPT / protons (2D) (3D)
![Page 4: Intensity(modulated/radiotherapy/ - Harvard University · 2016. 7. 13. · Overview/ Fluence!map!op.mizaon! Extensions!in!IMRT!planning! < Intensity](https://reader033.vdocument.in/reader033/viewer/2022052104/603ee721b6e171642b292bd8/html5/thumbnails/4.jpg)
Overview
Fluence map Dose distribu.on
di = x jj! Dij
![Page 5: Intensity(modulated/radiotherapy/ - Harvard University · 2016. 7. 13. · Overview/ Fluence!map!op.mizaon! Extensions!in!IMRT!planning! < Intensity](https://reader033.vdocument.in/reader033/viewer/2022052104/603ee721b6e171642b292bd8/html5/thumbnails/5.jpg)
Overview
1. Fluence map op.miza.on 2. Leaf sequencing
Two steps
Direct Aperture Op.miza.on (DAO)
Tradi-onal approach to IMRT planning:
Third component
![Page 6: Intensity(modulated/radiotherapy/ - Harvard University · 2016. 7. 13. · Overview/ Fluence!map!op.mizaon! Extensions!in!IMRT!planning! < Intensity](https://reader033.vdocument.in/reader033/viewer/2022052104/603ee721b6e171642b292bd8/html5/thumbnails/6.jpg)
Beam collimation using MLC
![Page 7: Intensity(modulated/radiotherapy/ - Harvard University · 2016. 7. 13. · Overview/ Fluence!map!op.mizaon! Extensions!in!IMRT!planning! < Intensity](https://reader033.vdocument.in/reader033/viewer/2022052104/603ee721b6e171642b292bd8/html5/thumbnails/7.jpg)
Beam collimation using MLC
![Page 8: Intensity(modulated/radiotherapy/ - Harvard University · 2016. 7. 13. · Overview/ Fluence!map!op.mizaon! Extensions!in!IMRT!planning! < Intensity](https://reader033.vdocument.in/reader033/viewer/2022052104/603ee721b6e171642b292bd8/html5/thumbnails/8.jpg)
Leaf sequencing
Superposi-on of mul-ple apertures yields an IMRT field
Leaf sequencing is the inverse problem
Find a set of MLC apertures that deliver an op.mized fluence map
![Page 9: Intensity(modulated/radiotherapy/ - Harvard University · 2016. 7. 13. · Overview/ Fluence!map!op.mizaon! Extensions!in!IMRT!planning! < Intensity](https://reader033.vdocument.in/reader033/viewer/2022052104/603ee721b6e171642b292bd8/html5/thumbnails/9.jpg)
Leaf sequencing
• Leaf sequencing does not have a unique solu.on
Goal: Find an efficient solu.on
(trivial solu.on: deliver each beamlet individually)
• reproduce fluence map faithfully • minimize number of apertures • minimize total number of monitor units
Construc-ve method: Sliding window (minimizes total MU)
Discrete op-miza-on methods: aim to minimize number of apertures
![Page 10: Intensity(modulated/radiotherapy/ - Harvard University · 2016. 7. 13. · Overview/ Fluence!map!op.mizaon! Extensions!in!IMRT!planning! < Intensity](https://reader033.vdocument.in/reader033/viewer/2022052104/603ee721b6e171642b292bd8/html5/thumbnails/10.jpg)
Sliding window sequencing
• leaves move uni-‐direc.onally
• leV leaf posi.ons:
• right leaf posi.ons: determined by posi.ve gradients
determined by nega.ve gradients
• consider discre.zed fluence map • consider one MLC leaf pair
• total MU = sum of posi.ve gradients
![Page 11: Intensity(modulated/radiotherapy/ - Harvard University · 2016. 7. 13. · Overview/ Fluence!map!op.mizaon! Extensions!in!IMRT!planning! < Intensity](https://reader033.vdocument.in/reader033/viewer/2022052104/603ee721b6e171642b292bd8/html5/thumbnails/11.jpg)
Direct aperture optimization
• Poor dose calcula.on accuracy for fluence map op.miza.on
Limita-ons of Two-‐step approach:
(use of pencil beam algorithm for Dose-‐influence matrix)
• Discrepancy between op.mized fluence map and sequenced map (treatment plan with few apertures)
• Inherent limita.ons of the dose-‐influence matrix concept (example: tongue & groove effect)
• discrete leaf posi.ons limited to beamlet boundaries (benefit of fine tuning leaf posi.ons at the target edge)
![Page 12: Intensity(modulated/radiotherapy/ - Harvard University · 2016. 7. 13. · Overview/ Fluence!map!op.mizaon! Extensions!in!IMRT!planning! < Intensity](https://reader033.vdocument.in/reader033/viewer/2022052104/603ee721b6e171642b292bd8/html5/thumbnails/12.jpg)
Direct aperture optimization
Tongue & Groove design of MLC leaves:
• dose-‐influence matrix assumes linearity
i.e. dose of beamlets delivered individually equals dose delivered by combined aperture
è center is underdosed if beamlets are delivered separately center is blocked as soon as one leaf is closed
• not strictly true
![Page 13: Intensity(modulated/radiotherapy/ - Harvard University · 2016. 7. 13. · Overview/ Fluence!map!op.mizaon! Extensions!in!IMRT!planning! < Intensity](https://reader033.vdocument.in/reader033/viewer/2022052104/603ee721b6e171642b292bd8/html5/thumbnails/13.jpg)
Direct aperture optimization
= directly op.mize shape and intensity of apertures Direct aperture Op-miza-on (DAO)
beamlet index (j) 1 2 3 4 5
3
2
1
5
4
leaf pair (n)
FMO DAO
![Page 14: Intensity(modulated/radiotherapy/ - Harvard University · 2016. 7. 13. · Overview/ Fluence!map!op.mizaon! Extensions!in!IMRT!planning! < Intensity](https://reader033.vdocument.in/reader033/viewer/2022052104/603ee721b6e171642b292bd8/html5/thumbnails/14.jpg)
Direct aperture optimization
Fluence map op-miza-on:
Dose is linear func.on of beamlet intensi.es
è efficient algorithms can find the global op.mum
![Page 15: Intensity(modulated/radiotherapy/ - Harvard University · 2016. 7. 13. · Overview/ Fluence!map!op.mizaon! Extensions!in!IMRT!planning! < Intensity](https://reader033.vdocument.in/reader033/viewer/2022052104/603ee721b6e171642b292bd8/html5/thumbnails/15.jpg)
Fluence map optimization
Objec.ves and constraints are ‘nice’ func.ons of the variables
xminimize wT x jDij
j! " 70#
$%%
&
'((
2
i)T! +wH xjDij
j!
i)H!
subject to x j ! 0 "jquadra.c func.on of xj
x jDijj! " 40 #i $ S
linear func.ons of xj
![Page 16: Intensity(modulated/radiotherapy/ - Harvard University · 2016. 7. 13. · Overview/ Fluence!map!op.mizaon! Extensions!in!IMRT!planning! < Intensity](https://reader033.vdocument.in/reader033/viewer/2022052104/603ee721b6e171642b292bd8/html5/thumbnails/16.jpg)
Direct aperture optimization
Fluence map op-miza-on:
Dose is linear func.on of beamlet intensi.es
è efficient algorithms can find the global op.mum
Direct aperture op-miza-on:
Dependence of dose on leaf posi.on is a smoothed step func.on
è highly non-‐convex op.miza.on problem with local minima
How many apertures should each beam direc.on get?
è combinatorial aspect
![Page 17: Intensity(modulated/radiotherapy/ - Harvard University · 2016. 7. 13. · Overview/ Fluence!map!op.mizaon! Extensions!in!IMRT!planning! < Intensity](https://reader033.vdocument.in/reader033/viewer/2022052104/603ee721b6e171642b292bd8/html5/thumbnails/17.jpg)
Direct aperture optimization
Three common approaches
• stochas.c search methods (simulated annealing)
• gradient-‐based leaf posi.on op.miza.on
• aperture genera.on methods
(commercialized by Prowess)
(RaySta.on, Pinnacle)
![Page 18: Intensity(modulated/radiotherapy/ - Harvard University · 2016. 7. 13. · Overview/ Fluence!map!op.mizaon! Extensions!in!IMRT!planning! < Intensity](https://reader033.vdocument.in/reader033/viewer/2022052104/603ee721b6e171642b292bd8/html5/thumbnails/18.jpg)
Gradient based DAO
di = yk! kni Lkn,Rkn( )
n"
k"
How does dose depend on leaf posi-ons?
dose in voxel i
sum over apertures k sum over MLC rows n
leaf/right leaf posi.ons
dose contribu.on of MLC row n in aperture k to voxel i
![Page 19: Intensity(modulated/radiotherapy/ - Harvard University · 2016. 7. 13. · Overview/ Fluence!map!op.mizaon! Extensions!in!IMRT!planning! < Intensity](https://reader033.vdocument.in/reader033/viewer/2022052104/603ee721b6e171642b292bd8/html5/thumbnails/19.jpg)
Dose dependence
• Assume that leO leaf is at the leO-‐most posi-on at the edge of the fluence map
Determine dose contribu-on of MLC row as a func-on of the right leaf posi-on:
!nki (Rnk )
![Page 20: Intensity(modulated/radiotherapy/ - Harvard University · 2016. 7. 13. · Overview/ Fluence!map!op.mizaon! Extensions!in!IMRT!planning! < Intensity](https://reader033.vdocument.in/reader033/viewer/2022052104/603ee721b6e171642b292bd8/html5/thumbnails/20.jpg)
Dose dependence
1 2 3 4
1
2
3
4
5
6
7
beamlet index j
leaf pair n
6 9 10 7 8 11 12 5 13 14
8
Loca.on of voxel i projected onto the MLC row
![Page 21: Intensity(modulated/radiotherapy/ - Harvard University · 2016. 7. 13. · Overview/ Fluence!map!op.mizaon! Extensions!in!IMRT!planning! < Intensity](https://reader033.vdocument.in/reader033/viewer/2022052104/603ee721b6e171642b292bd8/html5/thumbnails/21.jpg)
Dose dependence
Dijn
j=1
5
!
Loca.on of voxel i projected onto the MLC row
beamlet j / posi.on along the MLC row
Dose contribu.on !nki (Rnk )
Dijn
j=1
4
!
4!x 5!x
![Page 22: Intensity(modulated/radiotherapy/ - Harvard University · 2016. 7. 13. · Overview/ Fluence!map!op.mizaon! Extensions!in!IMRT!planning! < Intensity](https://reader033.vdocument.in/reader033/viewer/2022052104/603ee721b6e171642b292bd8/html5/thumbnails/22.jpg)
Dose dependence
Dose contribu-on of MLC row n:
! kni Lkn,Rkn( ) = !kni Rkn( )"!kni Lkn( )
contribu.ons from beamlets that are not blocked by the right leaf, assuming the leV leaf is at the edge of the field
subtract dose contribu.ons from beamlets blocked by the leV leaf
!nki (Rnk ) = piecewise linear func.on
(corners given by dose-‐influence matrix)
![Page 23: Intensity(modulated/radiotherapy/ - Harvard University · 2016. 7. 13. · Overview/ Fluence!map!op.mizaon! Extensions!in!IMRT!planning! < Intensity](https://reader033.vdocument.in/reader033/viewer/2022052104/603ee721b6e171642b292bd8/html5/thumbnails/23.jpg)
Direct aperture optimization
Treatment plan op.miza.on
xminimize wT di ! 70( )2
i"T# +wH di
i"H#
subject to x j ! 0 "j
minimize devia.on from 70 Gray in the tumor
fluence cannot be nega.ve
minimize dose in healthy .ssues
![Page 24: Intensity(modulated/radiotherapy/ - Harvard University · 2016. 7. 13. · Overview/ Fluence!map!op.mizaon! Extensions!in!IMRT!planning! < Intensity](https://reader033.vdocument.in/reader033/viewer/2022052104/603ee721b6e171642b292bd8/html5/thumbnails/24.jpg)
Direct aperture optimization
Treatment plan op.miza.on
xminimize wT di ! 70( )2
i"T# +wH di
i"H#
subject to x j ! 0 "j
minimize devia.on from 70 Gray in the tumor
fluence cannot be nega.ve
minimize dose in healthy .ssues
![Page 25: Intensity(modulated/radiotherapy/ - Harvard University · 2016. 7. 13. · Overview/ Fluence!map!op.mizaon! Extensions!in!IMRT!planning! < Intensity](https://reader033.vdocument.in/reader033/viewer/2022052104/603ee721b6e171642b292bd8/html5/thumbnails/25.jpg)
Direct aperture optimization
Treatment plan op.miza.on
y,L,Rminimize wT di ! 70( )2
i"T# +wH di
i"H#
(dose in voxel i) di = yk! kni Lkn,Rkn( )
n"
k"
Lkn ! Rkn "n,k (leaves cannot cross)
yk ! 0 "k (posi.ve aperture weights)
![Page 26: Intensity(modulated/radiotherapy/ - Harvard University · 2016. 7. 13. · Overview/ Fluence!map!op.mizaon! Extensions!in!IMRT!planning! < Intensity](https://reader033.vdocument.in/reader033/viewer/2022052104/603ee721b6e171642b292bd8/html5/thumbnails/26.jpg)
Gradient based optimization
Deriva-ve with respect to aperture weight:
di = yk! kni Lkn,Rkn( )
n"
k"
!f!yk
=!f!di
!di!yki
" =!f!di
# kni Lkn,Rkn( )
n"
$
%&
'
()
i"
![Page 27: Intensity(modulated/radiotherapy/ - Harvard University · 2016. 7. 13. · Overview/ Fluence!map!op.mizaon! Extensions!in!IMRT!planning! < Intensity](https://reader033.vdocument.in/reader033/viewer/2022052104/603ee721b6e171642b292bd8/html5/thumbnails/27.jpg)
Gradient based optimization
Deriva-ve with respect to leaf posi-ons:
di = yk! kni Lkn,Rkn( )
n"
k"
!f!Rkn
=!f!di
!di!Rkni
" =!f!di
yk!!kn
i Rkn( )!Rkn
#
$%
&
'(
i"
! kni Lkn,Rkn( ) = !kni Rkn( )"!kni Lkn( )
!Dijn
j = beamlet index where leaf edge is posi.oned
![Page 28: Intensity(modulated/radiotherapy/ - Harvard University · 2016. 7. 13. · Overview/ Fluence!map!op.mizaon! Extensions!in!IMRT!planning! < Intensity](https://reader033.vdocument.in/reader033/viewer/2022052104/603ee721b6e171642b292bd8/html5/thumbnails/28.jpg)
VMAT optimization
Volumetric modulated arc therapy (VMAT) • Con.nuous delivery mode:
Radia.on beam is constantly on while gantry and MLC leaves move
• Variables (conceptually):
• leaf posi.ons as a func.on of .me • gantry angle as a func.on of .me • dose rate as a func.on of .me
• Variables (in prac.ce): (driven by DICOM specifica.on)
Determine one aperture every 2 degrees
![Page 29: Intensity(modulated/radiotherapy/ - Harvard University · 2016. 7. 13. · Overview/ Fluence!map!op.mizaon! Extensions!in!IMRT!planning! < Intensity](https://reader033.vdocument.in/reader033/viewer/2022052104/603ee721b6e171642b292bd8/html5/thumbnails/29.jpg)
VMAT optimization
![Page 30: Intensity(modulated/radiotherapy/ - Harvard University · 2016. 7. 13. · Overview/ Fluence!map!op.mizaon! Extensions!in!IMRT!planning! < Intensity](https://reader033.vdocument.in/reader033/viewer/2022052104/603ee721b6e171642b292bd8/html5/thumbnails/30.jpg)
VMAT optimization
VMAT op-miza-on methods Ref: Unkelbach et al, Med Phys, 2015, ‘Op.miza.on approaches to volumetric modulated arc therapy planning’
• fluence map op.miza.on • (arc) sequencing • direct aperture op.miza.on
VMAT approaches reuse the concepts from IMRT planning
• Approaches differ in the exact implementa-on of each step, and on the step they rely on most
![Page 31: Intensity(modulated/radiotherapy/ - Harvard University · 2016. 7. 13. · Overview/ Fluence!map!op.mizaon! Extensions!in!IMRT!planning! < Intensity](https://reader033.vdocument.in/reader033/viewer/2022052104/603ee721b6e171642b292bd8/html5/thumbnails/31.jpg)
VMAT optimization
Example: Prostate case treated in a single 360 degree arc
Goal: • divide arc into 180 sectors of 2 degree lengths • assign one aperture (control point) to each sector • neighboring apertures should be similar
Raysta-on, Pinnacle, Monaco:
1. Fluence map op.miza.on 2. Arc sequencing 3. Direct aperture op.miza.on
Three-‐step approach (largely rely on DAO step)
![Page 32: Intensity(modulated/radiotherapy/ - Harvard University · 2016. 7. 13. · Overview/ Fluence!map!op.mizaon! Extensions!in!IMRT!planning! < Intensity](https://reader033.vdocument.in/reader033/viewer/2022052104/603ee721b6e171642b292bd8/html5/thumbnails/32.jpg)
VMAT optimization
Step 1: Perform fluence map op.miza.on at 20 equispaced angles
Step 2: approximate each fluence map through 9 apertures (yields 180 control points, one every 2 degrees)
typical approach: sliding window sequencing
Why?
![Page 33: Intensity(modulated/radiotherapy/ - Harvard University · 2016. 7. 13. · Overview/ Fluence!map!op.mizaon! Extensions!in!IMRT!planning! < Intensity](https://reader033.vdocument.in/reader033/viewer/2022052104/603ee721b6e171642b292bd8/html5/thumbnails/33.jpg)
VMAT optimization
Step 1: Perform fluence map op.miza.on at 20 equispaced angles
Step 2: approximate each fluence map through 9 apertures (yields 180 control points, one every 2 degrees)
typical approach: sliding window sequencing
Why? Naturally yields ordered apertures. Unidirec.onal leaf mo.on makes subsequent apertures similar.
![Page 34: Intensity(modulated/radiotherapy/ - Harvard University · 2016. 7. 13. · Overview/ Fluence!map!op.mizaon! Extensions!in!IMRT!planning! < Intensity](https://reader033.vdocument.in/reader033/viewer/2022052104/603ee721b6e171642b292bd8/html5/thumbnails/34.jpg)
Sliding window sequencing
![Page 35: Intensity(modulated/radiotherapy/ - Harvard University · 2016. 7. 13. · Overview/ Fluence!map!op.mizaon! Extensions!in!IMRT!planning! < Intensity](https://reader033.vdocument.in/reader033/viewer/2022052104/603ee721b6e171642b292bd8/html5/thumbnails/35.jpg)
VMAT optimization
Step 1: Perform fluence map op.miza.on at 20 equispaced angles
Step 2: approximate each fluence map through 9 apertures (yields 180 control points, one every 2 degrees)
typical approach: sliding window sequencing
Why? Naturally yields ordered apertures. Unidirec.onal leaf mo.on makes subsequent apertures similar.
Step 3: Perform gradient based DAO
one aperture at each of 180 gantry angles
![Page 36: Intensity(modulated/radiotherapy/ - Harvard University · 2016. 7. 13. · Overview/ Fluence!map!op.mizaon! Extensions!in!IMRT!planning! < Intensity](https://reader033.vdocument.in/reader033/viewer/2022052104/603ee721b6e171642b292bd8/html5/thumbnails/36.jpg)
VMAT optimization
DAO step yields a DICOM VMAT plan, specified by
for each control point:
• all leaf posi.ons • gantry angle • (couch and collimator angle) • total MU delivered up to this control point
Linac machine controller translates DICOM into trajectories
this does not contain -me!
(TPS relies on assump.ons to es.mate treatment .me)
![Page 37: Intensity(modulated/radiotherapy/ - Harvard University · 2016. 7. 13. · Overview/ Fluence!map!op.mizaon! Extensions!in!IMRT!planning! < Intensity](https://reader033.vdocument.in/reader033/viewer/2022052104/603ee721b6e171642b292bd8/html5/thumbnails/37.jpg)
VMAT optimization
How do aperture weights relate to gantry speed and dose rate?
yk =!k!ksk
aperture weight [MU]
angular distance between control points [degree]
dose rate [MU/s]
gantry speed [degree/s]
large aperture weight realized by high dose rate or low gantry speed
![Page 38: Intensity(modulated/radiotherapy/ - Harvard University · 2016. 7. 13. · Overview/ Fluence!map!op.mizaon! Extensions!in!IMRT!planning! < Intensity](https://reader033.vdocument.in/reader033/viewer/2022052104/603ee721b6e171642b292bd8/html5/thumbnails/38.jpg)
VMAT optimization
Constrain maximum leaf travel for efficiency:
assuming we want to deliver a 360 degree arc in one minute (maximum gantry speed)
gantry speed: s = 6° / second control point spacing: Δk = 2°
è ⅓ second per control point
maximum leaf speed: v = 6 cm / second
è maximum leaf travel between control points is Δ = 2 cm
![Page 39: Intensity(modulated/radiotherapy/ - Harvard University · 2016. 7. 13. · Overview/ Fluence!map!op.mizaon! Extensions!in!IMRT!planning! < Intensity](https://reader033.vdocument.in/reader033/viewer/2022052104/603ee721b6e171642b292bd8/html5/thumbnails/39.jpg)
Direct aperture optimization
y,L,Rminimize wT di ! 70( )2
i"T# +wH di
i"H#
(dose in voxel i) di = yk! kni Lkn,Rkn( )
n"
k"
Lkn ! Rkn "n,k (leaves cannot cross)
yk ! 0 "k (posi.ve aperture weights)
Rkn ! R(k+1)n " # (constrain maximum leaf travel between apertures)