Evolutionary Computing Enriched Computer-Aided Diagnosis System for Diabetic Retinopathy: A Survey

• Published in: IEEE reviews in biomedical engineering Vol. 10; pp. 334 - 349
• Main Author: Mansour, Romany F.
• Format: Journal Article
• Language: English
• Published: United States IEEE 2017; The Institute of Electrical and Electronics Engineers, Inc. (IEEE)
• Subjects: Ant colony optimization; Biomedical imaging; Blood vessels; Clustering; Complications; Computation; Computer-aided diagnosis (CAD); Design automation; Diabetes; Diabetes mellitus; Diabetic retinopathy; diabetic retinopathy (DR); Diagnosis; Evolutionary algorithms; evolutionary computing (EC); Exudates; Exudation; Feature extraction; Genetic algorithms; Hemorrhage; Hemorrhaging; image processing; Microvasculature; Particle swarm optimization; Retina; Retinopathy; Segmentation
• ISSN: 1937-3333; 1941-1189; 1941-1189
• DOI: 10.1109/RBME.2017.2705064

Complications caused due to diabetes mellitus result in significant microvasculature that eventually causes diabetic retinopathy (DR) that keeps on increasing with time, and eventually causes complete vision loss. Identifying subtle variations in morphological changes in retinal blood vessels, optic disk, exudates, microaneurysms, hemorrhage, etc., is complicated and requires a robust computer-aided diagnosis (CAD) system so as to enable earlier and efficient DR diagnosis practices. In the majority of the existing CAD systems, functional enhancements have been realized time and again to ensure accurate and efficient diagnosis of DR. In this survey paper, a number of existing literature presenting DR CAD systems are discussed and analyzed. Both traditional and varoius evolutionary approaches, including genetic algorithm, particle swarm optimization, ant colony optimization, bee colony optimization, etc., based DR CAD have also been studied and their respective efficiencies have been discussed. Our survey revealed that evolutionary computing methods can play a vital role for optimizing DR-CAD functional components, such as proprocessing by enhancing filters coefficient, segmentation by enriching clustering, feature extraction, feature selection, and dimensional reduction, as well as classification.