Deep Learning for Ophthalmology: The State-of-the-Art and Future Trends

TL;DR

This review surveys deep learning applications in ophthalmology across DR, glaucoma, AMD, and retinal vessel segmentation, emphasizing CNNs, attention mechanisms, and transformer-based models. It maps foundational DL methods to concrete clinical tasks, catalogs widely used public datasets, and highlights both performance gains and interpretability challenges. A core contribution is the synthesis of multimodal and transformer-based approaches, alongside a forward-looking discussion of XAI, longitudinal prognosis, and real-world deployment. The paper underscores the potential of AI to improve diagnostic accuracy and patient care while outlining practical requirements—data diversity, transparency, validation, and clinician collaboration—for safe, scalable clinical adoption.

Abstract

The emergence of artificial intelligence (AI), particularly deep learning (DL), has marked a new era in the realm of ophthalmology, offering transformative potential for the diagnosis and treatment of posterior segment eye diseases. This review explores the cutting-edge applications of DL across a range of ocular conditions, including diabetic retinopathy, glaucoma, age-related macular degeneration, and retinal vessel segmentation. We provide a comprehensive overview of foundational ML techniques and advanced DL architectures, such as CNNs, attention mechanisms, and transformer-based models, highlighting the evolving role of AI in enhancing diagnostic accuracy, optimizing treatment strategies, and improving overall patient care. Additionally, we present key challenges in integrating AI solutions into clinical practice, including ensuring data diversity, improving algorithm transparency, and effectively leveraging multimodal data. This review emphasizes AI's potential to improve disease diagnosis and enhance patient care while stressing the importance of collaborative efforts to overcome these barriers and fully harness AI's impact in advancing eye care.

Figures (10)

• Figure 1: Overall schematic diagram describing four main problems in common ophthalmic imaging modalities presented in our survey.
• Figure 2: Rapid growth in publications leveraging machine learning and deep learning for ophthalmology from 2012 to 2023.
• Figure 3: Overview of our survey, highlighting application categories and key methods are used.
• Figure 4: Proposed AI-human hybrid workflow: AI-screened fundus images labeled as more-than-mild diabetic retinopathy (MTMDR)-positive or AI-ungradable are overread by a human expert in teleophthalmology. Patients with an MTMDR-negative outcome undergo AI rescreening in 12 months, while those with an MTMDR-positive result or ungradable images are referred for in-person examination dow2023ai.
• Figure 5: Overall architecture of deep learning-based Computer-Aided Diagnosis for diabetic retinopathy detection son2023interpretable.