Trial lecture
Medical image synthesis using generative AI: Applications in ultrasound such as denoising, super-resolution and domain transfer
Time and place: January 16, 2026 11:15 AM. Kristen Nygaards sal (5370), Ole-Johan Dahls hus.
Main research findings
My research presents new methods to create smoother, high-quality videos of the heart using ultrasound imaging, without sacrificing image sharpness. These improved videos provide better visualization of heart motion and smoother display of cardiac structures.
Ultrasound imaging faces a fundamental challenge: to capture clear images of the rapidly moving heart, doctors must choose between smooth videos or sharp images. Getting both at once has been difficult because of physical limitations in how sound travels through the body.
We developed three smart algorithms that fill in missing frames between captured images. Think of it like creating smooth slow-motion video from a handful of still photos. Our methods track how different parts of the heart move and use this information to generate realistic in-between frames. The key innovation is making these methods fast enough for real-time use in hospitals and robust enough to handle the noisy nature of ultrasound images.
We also showed that artificial intelligence can learn from our methods to work even faster. This breakthrough makes it practical to use these techniques during actual patient examinations, enabling smoother visualization of heart motion in real-time.
Adjudication committee
- Professor Elsa Angelini, Telecom Paris, France
- Associate Professor Gabriel Hanssen Kiss, NTNU, Norway
- Professor Andreas Austeng, Department of Informatics, University of Oslo, Norway
Supervisors
Professor Anne Helene Schistad Solberg, University of Oslo, Norway
Professor Martin Reimers, Fridtjof Eiendom AS, Norway
Principal Engineer Erik Norman Steen, GE Healthcare, Norway
Professor Emeritus Sverre Holm, University of Oslo, Norway
Ph.D. Sten Roar Snare, GE Healthcare, Norway
Chair of defence
- Associate Professor Aparecido Fabiano Pinatti de Carvalho, Department of Informatics, University of Oslo
Contact information at Department: Mozhdeh Sheibani Harat