Optimized Convolutional Neural Network based Multiple Eye Disease Detection and Information Sharing System

Abstract

Detection of various eye diseases using a multi-label classification technique is an efficient way. The key benefit of this technology is that it can detect diseases at an initial stage. In this paper, a deep convolutional neural network (DCNN) was used to detect various eye diseases, namely, cataracts, glaucoma, DR, and age-related macular degeneration. For optimizing various hyperparameters of DCNN, a whale optimization algorithm was used. The retinal fundus images from the different databases have been used in this study 75% of datasets were used for the initial training and the remaining were used for testing. By using the proposed optimization technique an improvement in the accuracy of 8.1% was obtained. Here a multiclass support vector machine (MSVM) was used for the classification of the above-mentioned diseases from the output of DCNN. Various performance measures (accuracy, specificity, sensitivity, precision, and F1-score) have been used to study the performance of the proposed optimized DCNN. The result showed that the detection accuracy of cataracts was 96.4%, for glaucoma 97.2%, for DR 97.4%, and for AMD 97.7% were obtained for the optimized DCNN.