Medical Image Analysis

The Medical Image Analysis team develops deep learning and computer vision methods that turn medical images into reliable, patient-specific information for clinicians. Our core research covers image segmentation (including topology-aware segmentation of vascular structures), deformable image registration, shape reconstruction, and geometric representations of anatomy such as point clouds, meshes and implicit neural representations. We also work on multimodal learning that combines imaging with clinical, tabular and textual data. We work across a wide range of modalities, including CT, MRI, CBCT, ultrasound, digital pathology and microscopy.

Our research is driven by concrete clinical problems. Examples include the automatic design of personalised implants, non-invasive diagnosis of pulmonary hypertension, AI-based mapping of glioblastoma infiltration, and digital twins of the facial vasculature. The team regularly takes part in international challenges at MICCAI and ISBI, where its methods have won several first places. We work closely with clinical partners and we keep our research open and reproducible whenever possible.