Institute of Artificial Intelligence · Medical University of Vienna

Advancing healthcare through intelligent medical image analysis

We build machine learning methods for retinal imaging — to measure what clinicians currently estimate by eye, and to predict how disease will progress in an individual patient.

Retinal disease is the leading cause of blindness, and optical coherence tomography has become the standard way of looking at it. A single OCT volume holds far more information than can be read off it in a routine appointment. Our methods extract that information automatically: the thickness of individual retinal layers, the volume of fluid driving a treatment decision, the extent of an atrophic lesion and the rate at which it is growing.

Deep learning matched human pattern recognition on these tasks some years ago. Getting it into clinical use is the harder problem, and it is the one we work on — models that hold up when the scanner changes, that learn from the mostly unlabelled data a hospital actually has, and that say when they are unsure.

We sit in the Institute of Artificial Intelligence at MedUni Vienna and work closely with clinical partners in ophthalmology. That is where the questions come from, and where the answers have to hold up.

Explore our research →