Retinal blood vessels extraction using morphological operations

Abstract

Automatic extraction of retinal blood vessels is an important task for computer aided diagnosis from retinal images. Without extracting of blood vessels, the structures with pathological findings such as microaneurysms, haemorrhages or neovascularisations could be erroneously exchanged. We developed two independent methods; every method is the combination of different morphological operations with different structural elements (different types and sizes). Images from standard database with blood vessels marked by ophthalmologist were used for evaluation. Sensitivity, specificity and accuracy were used as measures of methods efficiency. Both approaches show promising results and could be used as a part of image preprocessing before pathological retinal findings detection algorithms.