An Ensemble CNN Model for Identification of Diabetic Retinopathy Eye Disease

Abstract

Diabetic retinopathy is an eye disease which occurs due to the impairment of the blood vessels at the retinal sensitive tissues. The disease can become severe if proper identification and management are not handled through comprehensive eye examination of ophthalmologists. Several techniques combined have been proposed using classification methods, machine learning algorithms, and require more time for training and testing. Also, the algorithms are not validated on different datasets. Hence, this paper is emphasized on the development of a robust and efficient classification model of ensemble CNN to identify the prevalence of diabetic retinopathy from the retinal images and suggest appropriate interventions at an early stage. The basic CNN and ensemble of CNN are implemented for classification of diabetic retinopathy and are compared with accuracy for best performance. The outcome of results depicts that the ensemble method is more accurate than a basic CNN due to the selection of hyper-parameters and obtained an accuracy of 75%. Therefore, it is ascertained that the ensemble CNN outperforms other methods of classification for identification of diabetic retinopathy.