Towards optimal stopping in radiation therapy

Abstract

A typical fractionated radiotherapy (RT) course is a long and arduous process, demanding significant financial, physical, and mental commitments from patients. Each additional session of RT significantly increases the physical and psychological burden on patients and leads to higher radiation exposure in organs-at-risk (OAR), while, in some cases, the therapeutic benefits might not be high enough to justify the risks. Today, through technological advancements in molecular biology, imaging, and genetics more information is gathered about individual patient response before, during, and after the treatment. we highlight some of the ways that mathematical tools can help assess treatment efficacy on the fly, adapt the treatment plan based on individual biological response, and optimally stop the treatment, if necessary. We term this "Optimal Stopping in RT (OSRT)", after a similar concept in the fields of dynamic programming and Markov decision processes. In short, OSRT can dynamically determine "whether, when and how" to stop the treatment to improve therapeutic ratios.

Keywords: Adaptive radiotherapy; Bayesian learning; Mathematical optimization; OSRT; Optimal stopping; Predictive biomarker.