Our research group aims to improve patient outcomes by applying artificial intelligence (AI) and data science to clinical data. A particular focus of our group is using these technologies for the analysis of medical images. AI and data science have the potential to transform medicine and improve patients outcomes in several ways:

● Making medical data analysis faster and more accessible, for example by partially automating diagnosis, outcome prediction, image quantification, and reconstruction.
● Creating entirely new clinical workflows that would not be feasible without AI support.
● Extracting new insights from large datasets to better guide treatment decisions, clinical research, and future drug development.

While AI technology is now ubiquitous in our daily lives in areas like voice assistance, online content moderation or fraud detection, its adoption in clinical practice remains surprisingly limited. One reason is that healthcare is a high-stakes environment where algorithms must meet exceptionally high standards of reliability and robustness. Another key challenge is the black-box nature of modern neural network technology. Clinicians may be reluctant to rely on systems whose decisions they cannot interpret or explain to patients. Both of these concerns not only affect clinical uptake but also play a central role in the regulatory and certification processes for medical AI.

Therefore, in order to start harnessing the massive potential of machine learning for healthcare, and to actually use it to improve real patient outcomes, our research group aims to bridge this gap between machine learning and clinical practice. We perform this research along five broad directions:

● Robustness, safety and uncertainty
● Interpretable machine learning
● Human-in-the-loop machine learning systems
● Generative modelling on big medical datasets
● Meta learning, and learning from small medical datasets