Developing a Predictive Model Using MRI Radiomics for Online Adaptation Strategies in MRI-Guided Radiotherapy for Prostate Cancer

Motivation: Changes in inter-fractional MRI radiomics can reveal some information to assist the objective online MRI-guided RT (MRgRT) plan adaptation. Goal(s): The primary goal is to develop an MRI radiomics model that could assist in determining the appropriate adaptation strategy based on daily-MRI images Approach: Data from 53 patients undergoing MRgRT were analyzed, extracting 1023 radiomics features. A machine learning model that incorporated reliable change index was built to determine adaptation strategies, and validated through independent testing sets. Results: The model performed well, and demonstrated that MRI radiomics has the potential to assist online MRgRT plan adaptation quantitatively and objectively in localized PC patients. Impact: The results of this study could assist clinicians' decision-making in MRI-guided radiotherapy (MRgRT) for localized prostate cancer, enhancing treatment precision and minimizing side effects. Building MRI radiomics models demonstrates the possibility for personalized treatment plans, ultimately improving patient outcomes.