MENU

Click here to

×

Are you sure ?

Yes, do it No, cancel

Fully Probabilistic Optimization Framework That Integrates Uncertainty in Target Volume Definition

G Buti1*, K Souris1, A M Barragan Montero1, J A Lee1, E Sterpin1,2, (1) Molecular Imaging, Radiotherapy & Oncology, Universite Catholique de Louvain, Brussels, BE Belgium, (2) Laboratory of Experimental Radiotherapy, Katholieke Universiteit Leuven, Leuven, BE Belgium

Presentations

(Sunday, 7/12/2020)   [Eastern Time (GMT-4)]

Room: AAPM ePoster Library

Purpose: Uncertainty in target volume definition is a crucial aspect in current radiation therapy clinical practice. This works aims to include this uncertainty, together with patient setup errors, in a fully probabilistic robust optimization scheme.

Methods: Setup errors and tumor infiltrations are considered sources of geometric and target volume uncertainty, respectively. The probabilities of the aforementioned uncertainties are represented by Gaussian probability functions with parameters from literature.
A 3D-tumor probability map is constructed by simulating realizations of the target using a Monte-Carlo sampling of the tumor infiltration. For each sampling, the GTV is dilated isotropically (corrected for anatomical barriers) and the value of each voxel is incremented if the target is realized there. This tumor probability map is fed to a probabilistic optimization algorithm that uses the obtained probability weights to calculate the expected value of the objective function.
The proposed optimization strategy was implemented in the open-source robust optimizer MIROpt and benchmarked against a conventional worst-case robust optimization. An intensity-modulated-proton-therapy plan was created using both methods, for a phantom case with a GTV (60Gy) and a nearby organ-at-risk (OAR). Plan robustness was evaluated with the MCsquare dose engine by recalculating the dose distribution on a set of 100 randomly sampled evaluation scenarios. Each evaluation scenario is composed of a setup error realization and microscopic tumor infiltration realization, sampled from the same distributions used in the optimization process.

Results: The fully probabilistic optimization method achieves a worst-case CTV D95 = 57.15 Gy and nominal OAR V15 = 10.0%, whilst the minimax plan gives worse target coverage and increased OAR irradiation: worst-case CTV D95 = 54.8 Gy and nominal OAR V15 = 21.8%.

Conclusion: A fully probabilistic optimization framework is implemented and achieves promising results for a phantom case by finding a better compromise between robust target coverage and OAR sparing.

Keywords

Protons, Optimization

Taxonomy

Not Applicable / None Entered.

Contact Email