ESCRS - PO648 - Optimized Ai-Based Early Detection Of Glaucoma And Cataracts Using Inceptionv3 And The Gc-Eyenet Dataset

Optimized Ai-Based Early Detection Of Glaucoma And Cataracts Using Inceptionv3 And The Gc-Eyenet Dataset

Published 2025 - 43rd Congress of the ESCRS

Reference: PO648 | Type: Free paper | DOI: 10.82333/fp9h-e933

Authors: Chang Won Park* 1 , Tae Young Gil 2 , Jae Pil Jeong 3 , Hyun Uk Park 4 , Ying jun Li 5

1Department of Optometry,Baekseok Culture University,Cheonan,Korea, Republic Of, 2Ophthalmology Clinic,Min Eye Clinic ,Cheongju,Korea, Republic Of, 3Department of Art & Technology ,Chung-Ang University,Anseong,Korea, Republic Of, 4Department of leisure and sports,Baekseok Culture University,Cheonan,Korea, Republic Of, 5Department of Ophthalmology ,FuyangPeople’s Hospital of Anhui Medical University,Anhui,China

Purpose

The purpose of this study is to develop an advanced AI-driven approach for the early detection of glaucoma and cataracts, two prevalent ocular diseases that can lead to irreversible vision loss if not diagnosed early.

Setting

This study utilizes the GC-EYENET dataset, a comprehensive collection of ocular images curated specifically for glaucoma and cataract detection, validated by ophthalmologists to ensure accuracy and reliability.

Methods

We combine deep learning techniques, specifically InceptionV3, with nature-inspired metaheuristic optimization strategies, including Chain Foraging and Cyclone Aging. These methods enhance the hierarchical feature extraction of InceptionV3, optimizing its diagnostic performance for glaucoma and cataracts.

Results

Experimental results show that the InceptionV3 model, optimized with these strategies, outperforms traditional machine learning methods. The model achieved an accuracy of 92.11%, with a precision of 92% and a recall of 90%, demonstrating significant improvement in early detection of both diseases.

Conclusions

Our AI-driven approach provides an efficient, reliable, and automated tool for the early detection of glaucoma and cataracts. This method improves diagnostic accuracy and reduces misclassification risks, making it beneficial for clinical applications and resource-limited environments. The study lays the foundation for further research in AI-powered ophthalmic diagnostics, with the potential to revolutionize early detection practices globally.