Generative Modeling of Anatomy for Quantification and Medical Image Segmentation – AnatomIA

Anatomical digital twins derived from medical images are key enablers of computer-aided diagnosis or prognosis, treatment planning or subtype discovery. We posit that recent advances in generative AI (diffusion models, LLM-like models) will bring about significant benefits towards better image-based anatomical models, with use cases for population-level and subject-level quantification and predictions. To be precise, the main investigation will be on developing new methodologies and paradigms, based on state-of-the-art generative AI, for two related challenges: 1) image-based generative modeling of anatomy; 2) anatomy-informed and data-efficient medical image segmentation.