Image-guided learning of dose distributions
A quick plan adaptation requires a dose computation in almost real-time, which is the topic of Research Project 1.2: Image-guided learning of dose distributions. This project makes use of deep-learning (DL) for dose computations. While DL techniques have become widespread in research, development and commercial products
regarding radiation therapy, DL algorithms have been subject to improvements, too. The project will employ a novel approach, which is well-suited for the transport of fast protons. Training data will comprise patient data from the WPE and novel imaging modalities such as counting-mode Spectral CT. In the second funding period, physics data will serve as an ingredient in the DL code. To summarize, Research Projects 1.1 and 1.2 will provide key instrumentation and algorithmic methods to pave the way to a high-precision PT by online, adaptive proton therapy.
