Abstract: There has been a growing interest in personalized radiation therapy (RT) in recent years. In this talk, I will present our ongoing work on personalized treatment planning for non-small cell lung cancer (NSCLC) patients. The proposed framework utilizes patient-specific data and dosimetric information to design an RT treatment plan optimization model with adverse side-effects constraints directly learned from historical and patient-specific data. We trained several predictive models to estimate the probability of grade 2+ radiation pneumonitis (RP) post-RT and embedded the best-performing model in the optimization process to constrain the patient-specific RP outcome below a pre-specified threshold. The framework was tested on different patients to optimize the dose distribution based on the patient's estimated radiosensitivity, and the pre- and post-adaptation plans were compared based on the difference in dosimetric profile and predicted RP risk.

About the speaker: Donato Maragno is a third-year PhD student at the Department of Business Analytics, University of Amsterdam (Netherlands). His research interests focus on the investigation of different techniques to embed Machine Learning into optimization models. He is one of the developers of OptiCL, an open-source tool for optimization with constraint learning.