LEVERAGING CONVOLUTIONAL NEURAL NETWORK TECHNIQUE FOR GLAUCOMA DETECTION IN RETINAL FUNDUS IMAGES

Electronics and Communication Engineering

Glaucoma, a pernicious and intractable chronic ocular pathology resulting in a persistent and irreparable visual debilitation, presents a ubiquitous and vexing global health predicament owing to its overwhelming prevalence as the paramount cause of ocular morbidity. The diagnostic process is a convoluted and laborious undertaking that demands a high degree of clinical expertise and interpretive acumen. Nevertheless, the advent of a deep learning-based automated diagnostic mechanism proffers a highly auspicious solution to surmount the intricate diagnostic conundrum. A meticulously curated Convolutional Neural Network (CNN) model, incorporating the VGG19 architecture and meticulously trained with exacting rigor utilizing a voluminous and highly heterogeneous fundus image LAG dataset, preprocessed using advanced image processing techniques, yielded an exceptional performance, exhibiting a remarkably high level of accuracy that surpassed 92% and an extraordinary sensitivity of 97%. A comprehensive array of diverse and meticulously chosen metrics, including precision, F1 score, was employed to evaluate the model's efficacy, and a rigorous comparative analysis was carried out to authenticate its effectiveness, demonstrating highly auspicious prospects for its imminent practical clinical implementation.

This work is licensed under a Creative Commons Attribution-ShareAlike 4.0 International License.