Retinal Vascular Network Segmentation Using Adaptive Thresholding Method Based on LSRV

As digital retina imaging and automatic retinal vascular network analysis continue to find increasing usefulness in the field of biomedicine for the diagnosis, monitoring and management of various forms of human illness like hypertension, retinopathies, glaucoma and cardiovascular diseases, the mitigation of different complications such as nonhomogeneous illumination noise, vessel width variation and very low contrast of the small-width vessels in relation to the retinal fundus background, for an efficient segmentation performance remains a subject of on-going research. This paper investigates the use of an adaptive thresholding method based on local spatial relational variance (LSRV) for the segmentation of the retinal vascular networks in fundus images. An experimental study conducted on DRIVE database shows that the vascular network segmentation results obtained from the investigated method detects large vessels and thin vessels in the retinal fundus images. When compared to some previous methods in the literature, the proposed method achieved higher average accuracy value of 95.04% and average sensitivity value of 76.55%. The proposed method is also computationally fast with a processing time of 4.5 seconds for the segmentation of the retinal vascular networks in each fundus image.