Computational Radiology

Computational radiology develops AI that reads and improves CT, MRI, X-ray and mammography — and that keeps working when it leaves the hospital where it was trained.

Our projects cover diagnosis, staging and treatment. PDAC-FL combines CT foundation models with heterogeneity-aware federated learning for multi-center staging of pancreatic cancer, and Prostate MRF detects clinically significant prostate cancer in 3D multiparametric MRI and MR fingerprinting. ChestAge estimates biological age from chest imaging, Fair Mammo builds bias-aware foundation models for equitable breast cancer screening, and LVM-Med pre-trains large vision models on 1.3 million medical images. With LMU Munich we built a deep learning toolkit for MRI-guided adaptive radiotherapy, and with Siemens Healthineers we lowered the X-ray dose of interventional imaging without losing image quality.

Federated learning (MSK-FL, FLamby) lets hospitals train together without sharing patient data, OMI brings AI inference into German university hospitals, and uncertainty estimation tells radiologists when not to trust a prediction.