Automated Cone and Vessel Analysis in Adaptive Optics Like Retinal Images for Clinical Diagnostics Support

Today, modern non-invasive Adaptive Optics (AO) imaging enables visualization of cone photoreceptors and vessels on a cellular level. High Magnification Module (HMM) images strongly resemble AO images and can be acquired fast and cost-effectly in clinical routine. Manual examination of those images, however, is tedious and time-consuming. Therefore, methods are needed to automatically analyse HMM images to facilitate the work of ophthalmologists. In this work an automatic cone detection method is presented that robustly detects cones in these images of both healthy and glaucoma patients. In addition, a vessel segmentation algorithm is provided to mask vessels during cone detection and additionally provide the ophthalmologist with vessel diameters that aid in monitoring ocular and cardiovascular diseases. The results on the given data are comparable to the performance of a trained expert and the methods are already being used in clinical practice.