Click here to


Are you sure ?

Yes, do it No, cancel

Online Plan Adaptation of Head and Neck IMPT Treatments Based On Cone Beam CT Imaging and GPU Monte Carlo Simulations

P Botas1,2*, J Kim4 , B Winey1,3 , H Paganetti1,3 , (1) Massachusetts General Hospital, Boston, USA, (2) Heidelberg University, Heidelberg, Germany, (3) Harvard Medical School, Boston, USA, (4) Yonsei Caner Center, Seoul, South Korea


(Wednesday, 8/1/2018) 1:45 PM - 3:45 PM

Room: Davidson Ballroom A

Purpose: To demonstrate an online IMPT plan adaptation algorithm based on GPU Monte Carlo (gPMC) and cone beam CT (CBCT) imaging to improve the daily dose distribution aiming to recover the original plan quality.

Methods: Replicating an adaptive workflow, IMPT plans of 5 patients were evaluated with gPMC on 5-7 weekly scatter-corrected CBCTs. Contours were propagated to the weekly geometries with a vector field (VF) calculated with deformable image registration between the CT and the CBCTs. First, beamlets were shifted following the VF at their distal falloff and raytraced in the weekly CBCT to adjust their energies, creating a geometrically adapted plan. Three geometric spot adaptation modes were considered: unconstrained geometric shifts, an isocenter shift, or a range shifter were employed. Second, for each mode, the weights of a selected subset of beamlets were automatically adjusted to match planned DVHs using gPMC-generated dose-influence matrices. All 6 adaptations were verified on the CBCTs and scaled to CTV V98=98% of 60 Gy(RBE) plans.

Results: The geometrical adaptation alone only worked with small anatomy changes. The weight-tuned adaptation worked for all 30 fractions, with the isocenter and unconstrained modes performing similarly (no statistically significant difference), superior to the range shifter. The calculation time per fraction was ~1 min. The CTV D2-D98 averaged across patients and fractions for the weight-tuned unconstrained method was 8.2 Gy(RBE), improved from the 14.1 Gy(RBE) of the unadapted plan on the CBCTs. The dose heterogeneity reduction ranged from 30.1 to 52.5% with respect to the unadapted plan. The maximum dose to mandible and larynx was reduced by 10.3% and 14.6% on average. Other organs at risk presented similar improvements.

Conclusion: An online MC-based adaptation algorithm was developed that significantly improved the plan quality in every studied fraction within clinically feasible time. The algorithm might allow planning margin reductions.


Not Applicable / None Entered.


Not Applicable / None Entered.

Contact Email