Identification of retinal diseases based on retinal blood vessel segmentation using Dagum PDF and feature-based machine learning

Retinal diseases is one of the major cause of vision impairments worldwide and the variations in fundus images help to identify these retinal diseases. Manual identification and segmentation of these fundus images is a tedious task and is highly prone to errors. Thus, a Dagum probability distribution function (PDF) based on matched filter (MF) for retinal blood vessel segmentation is proposed in this work. Initially, different features are extracted from the fundus images and these features are provided to the ensemble classifier to perform the task of disease classification. The overall performance of the proposed work is evaluated using different popular datasets such as DRIVE, STARE, HRF, IRDiD, Kaggle retinal, ORIGA and RFI. It is found that the specificity of 0.9802 and 0.9838 is achieved on DRIVE and STARE datasets respectively. In addition, the input image is predicted with HRF-glaucoma image '10_g.jpg', having a higher prediction accuracy of 95.45%.