A novel framework for retinal vessel segmentation using optimal improved frangi filter and adaptive weighted spatial FCM

• Published in: Computers in biology and medicine Vol. 147; p. 105770
• Main Authors: Mahapatra, Sakambhari, Agrawal, Sanjay, Mishro, Pranaba K., Pachori, Ram Bilas
• Format: Journal Article
• Language: English
• Published: Oxford Elsevier Ltd 01.08.2022; Elsevier Limited
• Subjects: AWSFCM; Blood vessels; Clustering; Diabetes; Diabetic retinopathy; ELPSO; Frangi filter; Genetic algorithms; Glaucoma; Image enhancement; Image segmentation; Inhomogeneity; Internal Medicine; Medical imaging; Methods; Morphology; Neural networks; Older people; Optimization techniques; Other; Parameter modification; Particle swarm optimization; Retina; Retinal images; Retinal vessel segmentation; AWSFCM; Retinal vessel segmentation; ELPSO; Frangi filter
• ISSN: 0010-4825; 1879-0534; 1879-0534
• DOI: 10.1016/j.compbiomed.2022.105770

Medical attention has long been focused on diagnosing diseases through retinal vasculature. However, due to the image intensity inhomogeneity and retinal vessel thickness variability, segmenting the vessels from retinal images is still a tough matter. In this paper, we suggest an optimal improved Frangi-based multi-scale filter for enhancement. The parameters of the Frangi filter are optimised using a modified enhanced leader particle swarm optimization (MELPSO). The enhanced image is segmented using a novel adaptive weighted spatial fuzzy c-means (AWSFCM) clustering technique. The suggested approach is tested on three freely available databases. The results obtained are compared with state-of-the-art procedures. It is observed that the suggested approach outperforms other methods and may serve as an effective approach for retinal vessel segmentation. •An improved Frangi vesselness function is optimised for vessel enhancement.•Modified ELPSO is suggested for optimization.•AWSFCM clustering is investigated for vessel segmentation.•Performance of optimised filter and AWSFCM fosters segmentation accuracy.•Comparison with state-of-the-art methods yields better results.