The aim of this project is to analyze and develop methods to mitigate the uncertainties affecting proton adaptive radiation therapy, in particular to investigate robust and probabilistic optimization and evaluation strategies for treatment planning. The uncertainties are related to multiple factors such as those involved in the daily routines for adaptive treatments (i.e., the algorithms dealing with the contouring of the daily patient image, the dose computation based on the lesser quality daily image, the accumulation of dose from previous fractions used as background dose for the daily treatment plan), or, on a more general level, to the  biological effects of the doses resulting from this type of planning in addition to the inherent uncertainties in such as the range of the protons, patient setup and breathing and organ motion.

For example, due to the extremely short time available for replanning, the contouring of the target and risk organs based on the daily image has to be done using some automated method, and the responsible physician will normally not have the opportunity to approve/modify the result prior to treatment planning and delivery. There are mainly two different methods for the automatic contouring: (1) transferring of the contours from the previous image using deformable registration, or (2) using some auto-contouring method, e.g. one based on machine learning. These two methods are inherently different and will exhibit different uncertainties. Previous studies performed by the group on incorporating relative biological effectiveness uncertainties into proton plan robustness evaluation showed, however, that the dominant factor with respect to uncertainties that need to be mitigated was the uncertainty in the radiobiological parameters describing the response of the tissue to radiation and therefore further consideration should be payed to it at the stage of treatment planning in the frame of robust optimization.

This work will consist of investigating the most appropriate optimization and evaluation strategies for the problem at hand, and to develop new techniques if needed, as well as quantifying the predicted uncertainties of the above methods by analyzing the underlying sources in detail. In addition, the clinical implications of accounting for these uncertainties will be investigated through comparative treatment planning studies for a portfolio of clinically relevant cases.