Inhibitor-Driven Reaction-Diffusion Algorithm for Retinal Image Segmentation and Blood Flow Visualization

This paper proposes a segmentation algorithm for retinal images by utilizing a reaction-diffusion system of the FitzHugh-Nagumo type mathematical model. The model describes propagation of an excited state along nerve axon by a time-evolving differential equation. The equation is tuned to a bi-stable system having two stable steady states, and its bi-stability is utilized for segmentation of a retinal image into two areas: blood vessel and its background. The algorithm is applied to sample images of open datasets, and it is found that strong inhibitory coupling drives segmentation process along blood vessels and the algorithm is also applicable to blood flow visualization.