Harnessing Deep Learning Methods for Detecting Different Retinal Diseases: A Multi-Categorical Classification Methodology

Abstract

Medical image classification plays a vital part in identifying and detecting diseases. Vision impairment affects 2.2 billion individuals globally, with cataracts, glaucoma, and diabetic retinopathy as major contributors. Timely diagnosis, crucial for effective treatment, often relies on imaging like color fundus photography. This study tackles multi-class classification challenges in retinal diseases using MobileNetV2. Traditional CNN models struggle with accuracy and efficiency, prompting the exploration of lightweight architectures. Leveraging MobileNetV2's efficiency, the aim is to improve diagnosis using a comprehensive ocular disease dataset. By integrating deep learning with conventional methods, growing challenges in ophthalmological analysis are addressed. The research underscores the importance of collaborative efforts in dataset curation, architecture design, and model interpretability to advance the multi-class classification of retinal diseases.