Retinal vascular topology estimation via dominant sets clustering

The estimation of vascular network topology in complex networks is important in understanding the relation between vascular changes and a wide spectrum of diseases. Automatic method of analysis of retinal vascular networks is of great assistance to the ophthalmologist in terms of diagnosis and treatment of eye disease. In this paper, we propose a method for estimating retinal vessel topology based on the concept of dominant sets clustering. Dominant sets clustering is a graph-theoretic approach that has proven to work well in data clustering, and has been successfully adapted to topology estimation in this work. The proposed approach has been applied to three public databases (IOSTAR, INSPIRE, and VICAVR) and achieved high accuracy of 0.915, 0.928, and 0.889 respectively. The experimental results show that it has effectively addressed crossover problem that is the bottleneck issue in reconstruct vascular topology. It is also worth noting that we have made manual annotations of vessel topologies from these databases, and these annotations will be released soon.