Artificial Intelligence and PET-Guided Proton Therapy

Over half of all cancer patients receive radiotherapy, in which radiation is delivered to tumors to eliminate cancer cells. Proton therapy, which uses high-energy proton beams, has emerged as a more precise alternative to conventional x-ray-based radiotherapy. However, treatment deviations lead to unwanted irradiation of healthy tissue, and the overall effectiveness of proton therapy could be improved with monitoring of tissue response. The radioactivity induced during proton therapy can be imaged using positron emission tomography (PET) and, combined with artificial intelligence and statistical modeling, may enable noninvasive detection of treatment deviations while providing biological insight into tissue response and vascular function. In our work, we develop tools for these purposes and demonstrate their application in a cohort of 23 patients imaged with PET following proton therapy at Massachusetts General Hospital. We highlight how advances in PET imaging, deep learning, and simulation have enabled these developments and discuss their potential to reduce healthy tissue toxicity and support biology-guided treatment.

Speaker: Pablo Cabrales Miró-Granada (RCCHU predoctoral fellow from Universidad Complutense de Madrid, visiting at Massachusetts General Hospital)