A self-adaptive optimisation for diabetic retinopathy detection with neural classification. (6th February 2020)

• Record Type: Journal Article
• Authors: Randive, Santosh Nagnath; Senapati, Ranjan K.; Rahulkar, Amol D.
• Is Part Of: International journal of nano and biomaterials. Volume 8:Number 3/4(2019)
• Page Start: 204
• Page End: 227
• Publication Date: 2020-02-06
• Subjects: diabetic retinopathy -- feature extraction -- classification -- self-adaptive grey wolf optimisation -- SA-GWO -- segmentation -- image processing -- feature selection -- neural network -- optimised training; Nanostructured materials -- Periodicals; Nanotechnology -- Periodicals; Biomedical materials -- Periodicals; 620.505
• Languages: English
• ISSNs: 1752-8933

Abstract

Diabetic retinopathy (DR) is a health issue which curtailed from diabetes for over long periods. When symptoms are rigorous, the patient might be blinded. The only way to avoid this problem is screening by an ophthalmologist. Anyhow, the automated detection utilises more time, and in addition, it includes several experts. Hence, this paper intends to propose a new diagnosis model of DR, which determines the severity of retinopathy. This model exploits three stages including segmentation, feature extraction, and classification. Here, density-based spatial clustering model (DBSCAN) is exploited for segmentation process, where it segments the abnormalities of retina. Coming to the feature extraction stage, this paper proceeds to extract the GLCM and GLRM features from the given input fundus image. Exclusively, it aims at a subjective contribution, i.e., the required features are optimally selected in this phase. Finally, for classification, this paper deploys a renowned neural network (NN) model. As the second contribution, the weight in NN model is optimally chosen. The optimal issues (selection of optimal features and weights) are successively solved by a new self-adaptive grey wolf optimisation (SA-GWO) approach. Moreover, the proposed method is evaluated with other conventional schemes, and the outcomes are obtained.