A Context for Effective Prediction and Classification of Diabetic Retinopathy Disease using Deep Ensemble AlexNet & LeNet Classifier

Typically, Diabetes is regarded as a metabolic disorder which aids in retinal complications termed Diabetic Retinopathy (DR) which is regarded as a major reason for blindness over globe. Usually, there exists no clear symptoms for DR before its onset, hence making disease identification an interesting job. Nowadays, industry of healthcare is facing negative significances once a gap in the identification of DR is not made with effectual automation. Hence, the objective of this work is to implement a cost-effective and automated model for the prediction and classification of DR samples. This research work presents a DL-based segmentation and classification of DR fundus images depending on its severity level. The proposed technique includes preprocessing, segmentation, feature extraction, optimal feature selection, and classification. The input retinal image dataset is considered and is subjected to preprocessing by filtering, restoration, augmentation, and enhancement process to remove noise and enhance the image quality. Then, segmentation process is carried with the use of Deep DenseNet-169 approach. The features from the segmented image are extracted by means of Enriched Haralick and SURF based feature point extraction technique. The optimal range of features are selected by means of Meta-heuristic tuna swarm optimization algorithm. At last, the classification process is carried by using Deep Ensemble AlexNet and LeNet algorithm. The classifier predicts and classifies the retinal image as per the disease severity. The disease was classified as Non-proliferative DR (NPDR), mild, and severe case. The performance assessment is made & the outcomes are then related with traditional models to show proposed model enhancement.