Automatic Detection of Maculopathy Using Deep Learning

Abstract

Diabetic maculopathy (DM) is a microvascular complication of diabetes that threatens central vision. Early detection is crucial for timely intervention and vision preservation. Traditionally, ophthalmologists diagnose DM through manual screening of retinal fundus images, a time-consuming process. This Work explores the potential of Deep Learning (DL) for automated DM stage detection. The main motive of the work been achieved by incorporating a trained Mobile Net on a dataset of fundus images Asia Pacific Tele-Ophthalmology Society (APTOS) on Kaggle containing five DM severity levels Mild, Moderate, Severe. The model extracts feature from these Fundus images and classifies them into the corresponding DM stages. The achieved accuracy of 0.96 demonstrates the promise of this approach for clinical applications. This work advances automated evaluation of diabetic ocular disorders using DL. Future initiatives for this research include assessing the model's efficacy in actual clinical situations and applying explainability methodologies.