Marta Hasny

Cardiovascular disorders and diseases are the leading cause of death worldwide, making them a critical area for deep learning applications. Medical foundation models have the potential to generalize across a multitude of tasks, while being trained in a self-supervised manner without the need for data annotation. These large deep learning models are trained on vast quantities of unlabeled data, resulting in models that can be adapted to a range of downstream tasks. Population database initiatives such as the German National Cohort (NAKO) and the UK Biobank have made vast amounts of medical data available to researchers, which can be utilized for training medical foundation models and comparing across German and UK populations. This doctoral project aims at the exploration of foundation models for cardiology, leveraging cardiac MRI data acquired in different spatial orientations, in cine mode, and longitudinally (follow-up imaging), as well as relevant tabular healthcare data.