ESTRO 2024 - Abstract Book

S75

Invited Speaker

ESTRO 2024

3444

Towards temporally-aware, biological radiotherapy planning

Ali Ajdari

Massachusetts General Hospital & Harvard Medical School, Radiation Oncology, Boston, USA

Abstract:

Radiotherapy (RT) is, by and large, a spatiotemporal treatment. Over the past few decades, significant advances have been made around spatial optimization of the dose: how to deliver a “sufficient” dose to the tumor while avoiding nearby healthy tissues. Current RT technologies such as intensity modulated proton and photon RT, stereotactic and radiosurgery techniques can deliver the radiation dose with sub-millimeter precision. Advanced optimization algorithms can deliver optimal modulation of the radiation field for many complicated, large-scale clinical cases. However, the temporal aspect of dose optimization has largely been neglected, leading to treatment plans that are designed to deliver the same treatment plan (dose) in every fraction. A major reason to deviate from this “temporally frozen” paradigm in favor of a more dynamic plan delivery is to account for the changing biological response of the patient over the course of treatment. Under a more temporally aware paradigm, one must determine how to optimally deliver the dose over time to maximize the therapeutic impact of RT. In this talk, we will discuss analytical and computational solutions for moving away from the current temporally frozen approaches towards a more dynamic treatment planning and delivery. The focus will be on how to translate potential biomarker signals gathered during the course of RT into informed treatment adaptation strategies. We will address the issues arising from biomarker uncertainties and discuss the emerging role of data-driven and machine learning powered optimization solutions for dynamic adaptation of the RT on the fly. Specifically, we will cover the following areas: • Dynamic biomarkers for response assessment and treatment adaptation : We will review some potential biomarkers for dynamic response assessment and RT plan adaptation, including those derived from imaging features, liquid biopsies, and patient-reported outcome measurements (PROMs). • Integration of artificial intelligence (AI) within RT planning: We will discuss some potential methods for translating dynamic biomarker signals into RT adaptation strategies, focusing on the emerging role of AI and data-driven optimization. The topic of data- and model uncertainties and some potential robust optimization solutions to address them will also be discussed. • Design of biomarker-informed clinical trials: Any significant deviation from the established clinical practice needs to be verified through robust clinical trials. We will discuss some potential solutions rooted in distributed and Bayesian clinical trial designs for testing temporally optimized RT strategies.