Comparative Analysis of Genetic Algorithm & Particle Swarm Optimization Techniques for SOFM Based Abnormal Retinal Image Classification

Genetic algorithm (GA) and particle swarm optimization (PSO) techniques have attracted considerable attention among heuristic optimization techniques. It is appropriate to compare their performance since both yield optimal solutions with different strategies and computational effort. In this paper, the application of these algorithms for feature selection in retinal image classification is explored. Abnormal retinal images from four classes namely non-proliferative diabetic retinopathy (NPDR), Central retinal vein occlusion (CRVO), Choroidal neo-vascularisation membrane (CNVO) and Central serous retinopathy (CRS) are used in this work. A comprehensive feature set is extracted from these images. An extensive feature selection is performed by the optimization techniques. The SOFM neural network is trained with optimal features and the classification accuracy is calculated. Experimental results show superior results for PSO optimized SOFM over the GA based SOFM in terms of classification accuracy and convergence time period. Index Terms Particle Swarm Optimization, Genetic Algorithm, Classification Accuracy, Retinal images.