Deep Learning Based Retinal Disease Classification Using Convolutional Neural Networks

Le résumé fourni par la source

The wide-reaching issue of vision impairment and blindness stems from various diseases affecting the retina, this strengthens the need for such ailments to be diagnosed early in their course. This research paper describes an entirely computer-aided technique for the classification of a few diseases affecting the retina on the basis of the analysis of several images of the retina using convolutional neural networks. The proposed model achieved a high accuracy of 96.8%, with a precision of 96.5%, recall of 96.3%, and an F1-score of 96.4%, demonstrating its effectiveness in detecting retinal abnormalities. The proposed method uses class distribution techniques and model parameters optimization to enhance performance of the algorithm. The investigations suggest that the model can generalize across the tests as instructed and performed in the controlled dataset but the generalizability to different clinical scenarios, understanding the predictions and computation still pose problems. As outlined, this work is intended to showcase the ability that artificial intelligence possesses in reshaping the traditional diagnosis of retinal diseases. If the remaining issues are solved, the algorithm can be further improved to aid in day to day clinical usage thereby increasing the speed, and accessibility to and efficiency of, diagnosis of diseases.

Ce résumé expose les affirmations des auteurs. BNTIC ne l’interprète pas comme une validation indépendante des résultats.