A novel automatic method for vessel tortuosity evaluation

Tortuosity evaluation of retinal or conjunctival vessels is one of the significant steps in early treatment of diabetic retinopathy. Despite the importance of this field, a few techniques have been proposed. In this paper, we proposed a new automatic algorithm for measuring vessel tortuosity based on Non Subsampled Contourlet Transform (NSCT). Major vessels and their directional information are extracted using NSCT in the first step. Then local vessel curvature is computed using obtained NSCT information and entire vessel network tortuosity is computed by combination of these local curvature values. Accuracy of our algorithm is evaluated by spearman correlation of our result and a set of images which are ordered by an ophthalmologist in ascending manner of tortuosity. We have shown that our algorithm achieves high accuracy in evaluation of vessels network tortuosity beside less computational time by reducing major steps of traditional tortuosity evaluation algorithm.