From position verification and correction to adaptive RT Adaptive RT and dose accumulation Hans de Boer Move away from Single pre-treatment scan Single treatment plan Treatment corrections by couch shifts only Standard margin recipes Adaptive radiotherapy (ART) Example: gradual changes H&N over course RT Term adaptive radiotherapy loosely defined Adapt to changing anatomy From very simple to very complex Anything beyond a single treatment plan and couch shifts In case of organ deformation, non-rigid registration and dose accumulation are useful tools for ART Just before radiotherapy Tumor and Sub-mandibular Glands Just after radiotherapy Parotid Glands Example: large deformation cervix Example: large deformation cervix Even with perfect translation (rigid body)corrections, margins so large (1-3 cm) that toxicity is dose limiting If deformation cannot be controlled a single treatment plan is not representative for delivered dose requires large margins Limited OAR sparing 729 cc 157 cc 729 cc GTV CTV Van de Bunt 2007 1
Concepts for ART Dose accumulation If deformation cannot be controlled Adapt dose to changing shapes Organ dose must be reconstructed Voxels need to be tracked from shape to shape Track volume elements during treatment Rely on non-rigid or deformable registration d 1 d? 2 cervix head and neck prostate Vásquez-Osorio 2009 organ fraction 1 organ fraction 2 Add dose per fraction d 1 +d 2 + (taking into account radiobiology) Dose accumulation Dose accumulation Dose important to biological effect is not N fractions D physics = d i i= 1 but for instance (LQ-model) NTD 2Gy = N fractions i= 1 α + di β d i α + 2Gy β Track volume elements during treatment to derive total dose Rely on non-rigid or deformable registration Retrospective Reconstruct delivered dose corrections and margins adequate? dose-effect relations Prospective Adapt dose during treatment: what is delivered dose? Take into account in constraints dose-effect relations Example retrospective dose accumulation: prostate and seminal vesicles Changing organ filling drives deformation Surface prostate and seminal vesicles Planning scan Week 6 of treatment 2
Residue errors: rotation and deformation Quantify deformation Plan on-line translation correction translation + rotation correction scans non-rigid registration (warping) treatment rotation deformation deformation vector fields Local deformation relative to markers Dose accumulation planned isodose lines planning scan organ Van der Wielen et al, 2008 standard deviation (mm) possible shapes and positions prostate and vesicles rectum (Mutanga et al, IJROBP 2010) Dose accumulation IMRT plan Delivered dose on-line corrections markers prostate margin 5 mm, vesicle 8 mm rectum vesicles prostate simulated DVH: accumulation over 35 fractions planned DVH Plan CTV Simulated CTVs Average (expected) DVH Simulated voxel position = mean position + local offsets (systematic and random) (Mutanga et al, IJROBP 2010) Prescribed 78 Gy 3
Example: margins derived for vesicles (not possible from margin recipe) Adaptive radiotherapy (ART) in practice From very simple to very complex Anything beyond a single treatment plan and couch shifts Adaptive radiotherapy: simplest Whole breast with (integrated) boost Adaptive radiotherapy off-line approach: Prostate Pre-treatment scan Initial plan 140 Start on initial scan planct Re-scan after two weeks Adapted plan Remaining fractions Initial plan on pre-treatment scan Acquire scans during first 4-6 fractions ( in-room CT, e.g. CBCT) Average position/shape target and OAR Replan and treat Take out systematic error in position AND shape 120 100 GTVvolume (cc) 80 60 40 fraction 1 1st boost fraction Yan 2000 Nuver 2006 20 0 0 20 40 60 80 100 Days since surgery PTV margin decrease from 1 cm to 7 mm Adaptive radiotherapy off-line: head and neck Adaptive radiotherapy: off-line Gradual volume changes: image repeatedly over treatment Example (Wu 2009) initial end There is a benefit but relatively small Parotids shift inwards (Barker 2004) Main dose effects is in parotids Wu 2009 original delineation Plan CT original delineation on CT week 6 new delineation on CT week 6 4
ART for highly deformable targets: Cervix ART on-line using off-line planning: library of plans Highly variable (random) anatomy: off-line predictions of systematic errors insufficient plan 1 plan 2 plan 3 plan 4 Adapt to anatomy of the day (moment)? T2-weighted MRI scans, Kerkhof 2008 bladder vol 157 cc 328 cc 586 cc 729 cc pre-treatment scans treatment fraction Use plan 1 Use interpolation plan 1 & 2 Luiza Bondar Med Phys 2010 250 cc smaller margins ART on-line using off-line planning: library of plans ART on-line using off-line planning: library of plans 37 ml 390 ml ITV = PTV for internal motion only PTV = ITV + extra margin Extra margin Setup errors Shape modelling errors! CTV shape not only determined by bladder filling! About 1 cm required Involved rectum volume (%) PTV = CTV + 24 mm CC, 24 LR, 36 AP full range ITV: patient specific margins 2 plans for 2 sub-range ITVs + margin 3 plans for 3 sub-rangeitvs + margin Bondar et al, IJROBP 2012 Very simple ART for cervix: Movers/non-movers ART with on-line (re-)planning: Cervix Based on pre-treatment scans with variable bladder filling Uterus tip motion < 1 cm Non-mover small margins Otherwise: mover conservative margins ctv Mover Non-mover Kerkhof UMCU 2008 5
ART with on-line (re-)planning: Cervix On-line and off-line plan creation Bowel volume with dose > 45 Gy Kerkhof UMCU 2008 Off-line creation: library of plans Create a library of plans based on pre-treatment scans and select best plan of the day based on imaging during treatment Assumptions: The driving variables for organ deformation can be varied during pre-treatment scanning The driving variables for organ deformation can be determined with treatment imaging Imaging, plan selection and delivery within time window for significant intrafraction motion Bladder, cervix (assuming bladder-filling as main driving force) Never a full adaptation to the anatomy of the moment On-line and off-line plan creation On-line/off-line ART requirements On-line ART: re-create plan on the spot Requirements Proper soft tissue contrast (MR-Linac) Fast recontouring (semi-automatic) Fast replanning and plan evaluation All in < 5 min for intrafraction motion Monitor intrafraction motion (MR-Linac) Basically applicable to any site Synergy with hypofractionation Requirements Automatic segmentation/contouring Automatic (validated) non-rigid registration Better image quality at treatment unit Dose accumulation Offline replanning/multiple planning: Very fast replanning/multiple planning available? early clinical early clinical MR linac early clinical clinical early clinical MRL prototype (Philips; Helsinki) ART requirement: auto-contouring ART requirement: auto-contouring cervix example green: automatic red : manual Full segmentation ~ minutes But what about manual editing 6
ART requirement: fast replanning Titel van de Slide Mostly segment adaptation, not complete replanning May become completely obsolete with GPU based methods! Segmented BEV Pre-tr CT Segmented BEV Fraction 7 deformable transformation initial fluence deformed fluence transform fluence Mohan, 2005 Monte Carlo calculations: highty accelerated in GPU hardware Accelerator head GPUMCD, fast GPU based Monte Carlo engine Monte Carlo code designed to run on GPU s (Hissoiny et al 2011) Benchmarked against EGSnrc and DPM Within 2%/2mm 900, resp 200 times faster using single NVidia GTX 480 A. Raaijmakers G. Bol S. Hissoiny C. Kontaxis B. Raaijmakers MRI linac: online treatment planning (Gijsbert Bol) FIDO IMRT optimization (Gijsbert Bol) Speed Using optimized Intel Math Kernel Library (MKL) Independent of magnetic field strength Online treatment planning MRI acquisition and segmentation Delivery 7
Nearly perfectly re-generated IMRT plan after 1 cm translation or 5 degrees rotation ART requirement: better visualisation MR linac examples 1 cm Translation (-5 Gy, +5 Gy) 5 Degrees rotation (-5 Gy, +5 Gy) DVH for Tumour (black), CTV (red), Bladder (turquoise) and Body (blue) Monaco is solid, MRLTP dashed ART summary Presently limited by availability of Soft tissue contrast on linac Commercially available tools for (fast, reliable) plan re-generation and delivery Sequential replanning for gradual changes Head and neck, breast, sometimes lung Library of plans for large day to day variations Plan of the day selection: mostly bladder related MR linac may boost this field to true adaptation (already happening in MR-guided brachytherapy) 8