Adaptive radiotherapy relies on accurate daily volumetric imaging to ensure precise dose delivery. While cone-beam CT (CBCT) and MRI are routinely used to monitor anatomical changes, they do not directly support reliable dose calculation. Deep learning–based synthetic CT (sCT) generation offers a promising alternative but is highly sensitive to variations in image quality introduced by hardware or software upgrades.

This PhD project focuses on the development of robust, AI-driven sCT models for adaptive proton therapy. The core of the work will involve deep learning method development, including transfer learning, fine-tuning strategies, and uncertainty-aware modeling to enable rapid revalidation of sCT models following imaging system changes. In parallel, the project will address medical physics aspects, evaluating how variations in sCT quality influence dose calculation accuracy, proton range robustness, and adaptive treatment workflows.

The project is embedded in a clinically relevant environment and offers interdisciplinary training at the interface of artificial intelligence, medical imaging, and radiation physics.