After radiology, ophthalmology is the field that generates the most image data. Consequently, there is a large amount of data that offers many opportunities for automated image data analysis. Research in Münster is mainly focused on the analysis of optical coherence tomography (OCT) data in patients with neovascular age-related macular degeneration (AMD). These patients are generally managed in our clinic according to a so-called Pro-Re-Nata scheme, i.e. they are seen every 4-6 weeks and receive an intravitreal injection if there are signs of activity. Signs of activity are specifically defined as the presence of intra- and/or subretinal fluid in the OCT.

In collaboration with Prof. Malcherek from the Department of Computer Science at the University of Applied Sciences in Darmstadt, Germany, we have developed deep learning algorithms that can accurately distinguish between active and inactive neovascular AMD. In another project, a deep learning algorithm was trained to automatically detect various changes typical of AMD in OCT images. Another focus is the automatic detection of glaucomatous optic nerve changes in glaucoma patients. An algorithm trained on fundus photographs achieved good statistical power in discriminating between optic nerves with glaucomatous damage and those with conspicuous changes but no nerve fiber layer loss. In this case, the algorithm performed as well as a glaucoma specialist when additional metadata were included. Further projects to establish methods for image annotation are in preparation.

• Prof. Julian Varghese, Institute for Medical Informatics, University of Münster
• Prof. Dr. Arnim Malcherek, Department of Computer Science, Darmstadt University, Germany