Speaker
Description
This talk presents a hybrid Artificial Intelligence – Monte Carlo (AI-MC) framework that integrates fast, data-driven radiation dose prediction with selective physics-based verification. In this approach, AI provides rapid full-volume dose estimates for real-time decision-making, while MC simulation is applied for targeted verification and refinement in clinically critical regions. MC simulation is the gold standard for radiotherapy dose calculation due to its high physical accuracy. But it is time-consuming and results in computational cost being too high to be feasible for time-sensitive computer-intensive clinical workflows such as image-guided adaptive radiotherapy. To overcome this limitation, AI can be used to approximate MC-generated dose distributions in seconds and enable rapid evaluation. But the shortcomings of AI still must be addressed. including generalizability, lack of explicit physical modeling, and uncertainty in out-of-distribution scenarios. In the talk, applications in adaptive radiotherapy and treatment planning will be discussed, demonstrating how hybrid methods can enable fast, accurate, and clinically robust dose calculation. By combining speed and accuracy, hybrid AI–MC methods offer a practical pathway toward real-time, physics-accurate radiotherapy.