Digital Twins for Radiation Oncology

Digital twin technology has revolutionized the state-of-the-art practice in many industries, and digital twins have a natural application to modeling cancer patients. By simulating patients at a more fundamental level than conventional machine learning models, digital twins can provide unique insights by predicting each patient's outcome trajectory. This has numerous associated benefits, including patient-specific clinical decision-making support and the potential for large-scale virtual clinical trials. Historically, it has not been feasible to use digital twin technology to model cancer patients because of the large number of variables that impact each patient's outcome trajectory, including genotypic, phenotypic, social, and environmental factors. However, the path to digital twins in radiation oncology is becoming possible due to recent progress, such as multiscale modeling techniques that estimate patient-specific cellular, molecular, and histological distributions, and modern cryptographic techniques that enable secure and efficient centralization of patient data across multiple institutions. With these and other future scientific advances, digital twins for radiation oncology will likely become feasible. This work discusses the likely generalized architecture of patient-specific digital twins and digital twin networks, as well as the benefits, existing barriers, and potential gateways to the application of digital twin technology in radiation oncology.