Beyond Digital Twins: Generative and Agentic Cardiac Modelling for Clinical Discovery
Cardiac imaging and clinical data are providing increasingly detailed views of heart structure, function, and disease progression. These data create new opportunities for personalised cardiovascular research, but they also expose a growing gap between what can be measured and what can be interpreted at scale. Current modelling pipelines often remain fragmented: imaging-derived measurements, clinical variables, population references, and downstream analyses are usually treated as separate steps, making it difficult to build coherent, reusable, and clinically meaningful workflows.
This talk explores how generative modelling and agentic workflows can support cardiac MRI analysis and multimodal clinical modelling. It focuses on learning population-level cardiac variation, defining personalised reference spaces, and supporting downstream tasks such as phenotype discovery, interpretation, and follow-up analysis. It also discusses key challenges, including clinical validity, multimodal and longitudinal data integration, workflow transparency, and the role of domain expertise in building reliable cardiac AI systems.
Short Bio
Dr Mengyun Qiao is a Lecturer at University College London, working at the intersection of artificial intelligence, cardiac MRI, and clinical data science. Her research focuses on generative modelling and multi-agent systems for understanding cardiac structure, function, and disease-related variation. Her work aims to develop scalable and interpretable approaches for cardiovascular research and personalised medicine. She has published widely in leading AI and medical imaging venues, including Nature Machine Intelligence, Nature Cardiovascular Research, IEEE Transactions on Medical Imaging, and MICCAI.