Trial lecture
“LLMs for advancing clinical care”
Time and place: December 18, 2025 11:15 AM. SmallTalk, Ole-Johan Dahls hus.
Main research findings
Aging populations mean that more people live with heart valve disease. Doctors often use echocardiography, a minimally invasive ultrasound examination of the heart, to diagnose and plan procedures. For the mitral valve, the supporting ring (the annulus) moves and changes shape throughout each heartbeat. Careful measurements are important for deciding if and how the mitral valve should be repaired. Automating these measurements can save time, improve consistency between observers, and support treatment decisions.
This thesis focuses on automating the tracing of the mitral valve ring in 3D ultrasound images. The methods use artificial intelligence that learns from real patient data. Over the course of the thesis, the models make increasing use of the full three-dimensional context in the images and more explicit knowledge about the ring’s typical saddle-like shape. In the experiments, the automatic measurements are on average within two millimetres of the reference annotations. This accuracy is similar to previously reported differences between human experts.
The results show that artificial intelligence can provide fast and consistent measurements of the mitral valve ring and could help cardiologists plan and guide valve repair procedures more efficiently.
Adjudication committee
- Professor Jan D'hooge, KU Leuven, Belgium
- Research Manager Ingerid Reinertsen, SINTEF Digital, Norway
- Professor Sven Peter Näsholm, University of Oslo, Norway
Supervisors
Professor II Eigil Samset,University of Oslo, Norway
Professor Anne Helene Schistad Solberg, University of Oslo, Norway
Associate Professor Sarina Thomas, University of Oslo, Norway
AI Researcher David Völgyes, Field Geospatial AS, Norway
Chair of defence
- Professor Carsten Griwodz, Department of Informatics, University of Oslo
Contact information at Department: Mozhdeh Sheibani Harat