Exploring the computational and clinical added-value of probabilistic optimization in proton therapy

Details

Supervisors

Lee, John Aldo ; Sterpin, Edmond

Faculty

Ecole polytechnique de Louvain

Degree label

Master [120] : ingénieur civil biomédical, à finalité spécialisée

Abstract

The definition of the clinical target volume can be considered as a binary mask. However, the further a point is from the gross target volume, the less probability it has of being tumorous. The probabilistic definition of the clinical target volume, called the clinical target distribution, is an effective alternative to adapt the dose distribution to the probability of tumor presence. This probabilistic definition allows one to maintain tumor control probability while reducing the dose delivered to the organs at risk. Modeling the cellular surviving fraction after irradiation can also be translated into cost functions for optimizations, yet studies so far that combine radiobiological cost functions and probabilistic definition of target volumes have been limited to radiobiological effects in the target. The spatial dependence of radiobiological optimization models makes them an excellent candidate for proton pencil beam scanning. In this Master thesis, we have developed a complete probabilistic radiobiological optimization framework for the target and organs at risk for proton pencil beam scanning. We have proven by the use of a head and neck clinical case that probabilistic radiobiological optimization allows for a decrease in the probability of normal tissue complications while maintaining tumor control probability comparable to classical dose-constrained radiobiological optimization methods.