Next Generation Deep Learning in Therapy Monitoring

In the Focus Group “Next Generation Deep Learning in Therapy Monitoring”, Hans Fischer Senior Fellow Anke Meyer-Baese (Florida State University, Scientific Computing) works with her hosts Prof. Claus Zimmer, Prof. Jan Stefan Kirschke and Prof. Benedikt Wiestler.

AI-assisted multimodal brain imaging has the potential to improve diagnostic accuracy by automatically delineating tumor boundaries and diagnosing tumors, predicting treatment outcomes and revolutionizing personalized medicine. Despite significant progress, challenges persist in early diagnosis and recurrence prediction of brain cancer. The latter is extremely difficult because of tumor heterogeneity, resistance to therapy and incomplete excision of cancer cells. The interpretability of AI-based results, as well as the asynchronous nature of follow-up data, requires the development of novel multi-step deep learning systems to obtain continuous tumor tracking. The goal is to develop an AI-powered treatment monitor for recurrence prediction and clinical decision support, focusing on the interpretability of the results to achieve the highest possible level of acceptance among physicians and patients. This will be achieved by developing a novel AI system combining the generative stochastic modeling with “physics-informed” neural identification algorithms that enable end-to-end tumor tracking using spatial and temporal graph neural attention network transformers.

 

TUM-IAS Funded Postdoctoral Researcher:
Dr. Sandeep Nagar, Image-Guided Diagnosis and Therapy