Understanding the eye through data — from single synapses to population-scale disease.

We analyze large and complex data sets in neuroscience and ophthalmology to advance our understanding of the healthy and diseased eye. To this end, we combine machine learning, mechanistic modeling and visualization with careful dataset curation and creation, working at the interface between data, domain knowledge and algorithms. Our methodological interests range from robust and interpretable models for clinical decision support to multimodal foundation models that link images with text, and from biophysical simulations and normative models of retinal circuits to diverse large-scale datasets of retinal structure and function.

With our clinical partners in the Department of Ophthalmology, we develop interpretable algorithms for improving clinical decision-making, detecting degenerative diseases of the eye early and modeling their time course. We work with fundus images, optical coherence tomography and large population-based cohorts such as NAKO and UK Biobank, focusing on major causes of vision impairment such as diabetic retinopathy and age-related macular degeneration. Beyond algorithm development, we build and curate large-scale ophthalmic datasets and benchmarks to support reproducible research in this domain, and we explore how language-grounded models can make medical image analysis more accessible. We are also committed to global health perspectives in ophthalmology, working with partners at the African Institute for Mathematical Sciences in Rwanda and the MRC Unit The Gambia to develop approaches that are effective beyond high-resource clinical settings.

In basic research, we work in close collaboration with experimental partners to integrate diverse data sources — including electrophysiological recordings, calcium imaging, single-cell transcriptomics and large-scale electron microscopy reconstructions — to link the computations performed by individual cell types to their genetics and anatomy. A particular focus is the retina, where we combine connectomic data with functional recordings to understand how circuit structure gives rise to neural computation and how these circuits are affected in degenerative diseases as part of the Eyewire II consortium. To support this work, we develop hybrid models incorporating mechanistic and statistical components as well as novel techniques for visual exploration of high-dimensional data.

We are part of the Cluster of Excellence "Machine Learning - New Perspectives for Science" funded in the Excellence Initiative and the BMBF Competence Center for Machine Learning Tübingen AI Center.