Reinforcement Learning in Oncology: A Comprehensive Review

Abstract

Reinforcement Learning (RL) is a framework that enables machines to learn dynamic decision-making strategies through trial and error, so as to maximize a numerical reward signal. Although its origins are rooted in early work across several disciplines, recent advancements have significantly enhanced its capabilities. Given the complexity and dynamic nature of cancer, RL holds significant promise in oncology, particularly in imaging, where it can enhance precision and efficiency. This paper presents a comprehensive review of RL in oncology, starting with a brief introduction to the basics of RL algorithms and their categories. It then provides an overview of various existing RL applications in oncology, including radiology and radiotherapy. The paper concludes with discussions on the current challenges and future perspectives, highlighting the potential of RL to transform cancer diagnosis and treatment through more personalized approaches.