CDBO_DenseNet: CAViaR Dung Beetle Optimization Enabled Deep Learning for Diabetic Retinopathy Classification

Diabetic Retinopathy (DR) occurs as a degrading infection that affects the eyes as well and it is an effect of Diabetes Mellitus (DM), where soaring levels of blood glucose provoke lesions on the eye retina. In this paper, CAViaR Dung Beetle Optimization-DenseNet (CDBO_DenseNet) is proposed for DR classification. At first, an image obtained from mentioned dataset is given towards a pre-processing module to enhance the image employing Mean Filter. Afterward, lesion segmentation is performed using Deep Joint Segmentation. Similarly, process of classifying artery and vein is performed using AVNet. Subsequently, feature extraction process extracts some essential features. Finally, DR is classified by DenseNet which is tuned with CDBO. CDBO is designed by integrating DBO with CAViaR technique. Additionally, CDBO _DenseNet accomplished high accuracy, sensitivity and specificity of 94 %, 92.9%, and 93.4% respectively.