Author(s): Pooja, Aman Kumar and Manvender Singh

Abstract: Glaucoma has been one of the major causes of permanent blindness in most parts of the world and early diagnosis is the most critical concern of ophthalmic care. It has emerged that recent developments in the field of machine learning (ML), deep learning methods such as convolutional neural networks (CNNs) in particular, are enormous in terms of the enhancement of diagnostic accuracy, efficiency, and availability during glaucoma screening. The given paper presents a complete review of 20 recent articles (20202025) to diagnose glaucoma based on various data types, such as fundus photos, optical coherence tomography (OCT), and outcomes of the visual field test. The results indicate that music playing ensemble models, multimodal learning, and explainable AI can improve predictive results and improve clinical confidence considerably. Moreover, AI solutions that can be moved demonstrate a possibility to overcome them in low-resource contexts. Even though such advances were made, the key research gaps were found in the homogenous datasets that limit generalizability, inability to integrate the research findings in the real-time clinical environment, and the absence of attention paid to the algorithm interpretability and ethical use. Future recommendations discussed in the review entail the creation of the variety of datasets, multimodal and longitudinal data, use of transparent AI structures and focus on collaboration between disciplines. These guidelines can be required to the translational of ML-based glaucoma detection systems into true clinical systems that will be used in everyday ocular management.

DOI: 10.33545/27075907.2025.v6.i2a.93

Pages: 07-14 | Views: 234 | Downloads: 118

Download Full Article: Click Here