Improved Invasive Weed Social Ski-Driver Optimization-Based Deep Convolution Neural Network for Diabetic Retinopathy Classification

Abstract

The eye-related problem of diabetes is called diabetic retinopathy (DR), which is the main factor contributing to visual loss. This research develops an enhanced deep model for DR classification. Here, deep convolutional neural network (Deep CNN) is trained with the improved invasive weed social ski-driver optimization (IISSDO), which is generated by fusing improved invasive weed optimization (IIWO) and social ski-driver (SSD). The IISSDO-based Deep CNN classifies DR severity into normal, mild, non-proliferative DR (NPDR), moderate NPDR, severe NPDR, and proliferative. Initially, a type 2 fuzzy and cuckoo search (T2FCS) filter performs pre-processing and the quality of the data is improved by data augmentation. The lesion is then divided using DeepJoint segmentation. Then, the Deep CNN determines the DR. The analysis uses the Indian DR image database. The IISSDO-based Deep CNN has the highest accuracy, sensitivity, and specificity of 96.566%, 96.773%, and 96.517%, respectively.