Boundary Extraction and Tortuosity Calculation in Retinal Fundus Images

Abstract

Tortuosity and retina blood vessel dilation are important symptoms of plus disease in retinopathy of prematurity. This paper presents differential geometrical method for tortuosity measurement. The problem of tortuosity evaluation is formulated as one-dimensional differential geometrical curvature characterization. The vessel network extracted from retinal image is subjected to boundary extraction and individual vessel boundaries are extracted as planar curves, further these curves are segmented and differential curvature is computed at segment level and at vessel level for the individual vessels. The method is tested and validated on the available public data and local data set. Vessels with considerable tortuosity are found to be having significant curvature variation compared to the normal vessels.