Intelligent model for diabetic retinopathy diagnosis: a hybridised approach

Abstract

As diabetic retinopathy (DR) is considered as most common infectious diseases in humans, more researches have been already proposed under various aspects, yet the attainment of accurate DR detection seems to be an issue. This paper intends to develop an innovative contribution by introducing a novel DR detection model, and further the proposed model tells the severity of retinopathy from the given input fundus image. The proposed model comprises of stages such as Segmentation, Feature Extraction and Classification. Here, Active contour model is used for segmentation; also the GLCM and GLRM features are extracted during feature extraction process. Moreover, the classifier called neural network (NN) is used for classification purpose. As a main contribution, the extracted features (feature selection), and weight in NN model are optimally chosen by a new hybridised algorithm called whale with particle swarm optimisation (WP), which compares its performance over other conventional methods for analysis purpose.