Explainability of Diabetic Retinopathy Detection & classification with Deep Learning Hybrid Architecture: AlterNet-k & ResNet-101

Diabetic Retinopathy(DR), eye disease that threaten cause of irreversible blindness. It always challenging to detect and diagnose early. There are several invasive procedures exists in Ophthalmology for diagnosis. All are required highly skilled medical practitioners with operational knowledge of diagnosing the sensitive organs like retina and its tiny vessels, due to dearth of retina specialist, eye’s organs sensitivity and complexity of retinal therapy the invasive procedure is time consuming, costly and have slow progress. The fundus images are the visual information of the rear part of the retina having progression of lesions around the retinal tissue’s surface, the electric signals not able to reach at visual cortex, then blurry vision or the vision loss experienced by the patients. The older methods of retinal fundus images for diagnosing lesions and symptoms of DR takes time, that causes delay in treatment and hence reducing the chance of success. Therefore, early diagnosis using fundus images can save the required efforts and time of doctors and patients. Artificial Intelligence (AI) techniques have the capability to learn the tissues structure of eye’s anatomy and to give the analysis about disease through the fundus images. Firstly, apply the image preprocessing techniques followed by splitting of dataset, creating multi head self-attention blocks, then classify the disease using the AI model. The proposed model should be trained over balanced dataset of DR images for prediction of accurate results followed by explain the decisions that diagnosed by model is correctly classified or not using Explainable AI algorithm.