Localizing Retinal Blood Vessels In Fundus Images using Fuzzy Logic Approach

The condition of the vascular system is crucial in the diagnosis of vision abnormalities. The essential element of computerized retinal analysis is vasculature segmentation in digital fundus images. This work introduces an automatic algorithm based on fuzzy logic to detect the retinal blood vessels. The methodology uses the green channel of images and employs pre-processing techniques. The features are extracted using the Robinson compass mask. The Mamdani interval type-2 fuzzy rules are applied to these features to detect the blood vessels, and the result is binarized, followed by post-processing refinements to maximize the performance. The methodology is evaluated using the publicly available DRIVE (Digital Retinal Images for Vessel Extraction) database, and the results are compared with distinguished published methods. Achieving the average accuracy of 94.89%, the sensitivity of 74.78% and specificity of 96.85% show promising results for pre-screening treatments while the technique could be extended to other image segmentation applications.