While machine learning approaches have become popular in recent years to investigate ophthalmic imaging of retinal structure, their utility to better understand visual function seems less obvious. Here, we introduce recent developments of artificial intelligence to improve the diagnosis of functional vision loss, to monitor its progression over time, and to relate retinal structure to specific visual impairments. Particularly for the disease of glaucoma, we demonstrate how machine learning made it possible to define and quantify functional disease subtypes and to identify subtype specific retinal biomarkers.

We perform longitudinal analysis of nAMD anti-VEGF Treatment Using Deep Learning. Specifically, we use Deep learning for semantic segmentation to investigate the influence of multiple clinical features in the retina on the treatment outcomes of patients with nAMD.

Using modern convolutional neural networks it is possible to detect systemic diseases - like peripheral artery disease. Furthermore our approach makes it possible to inspect the decision making process of the model using a novel attention mechanism. In this talk I will present the technical details of our approach, how well it performed and how it can improve diagnostics in ophthalmology.