Insight | Eye Discovery: How AI + multimodal data is redefining the ophthalmic diagnostic paradigm

The human eye is not merely an optical window to the world, but also a "micro-display" of systemic microcirculation and neural activity. In the current era of big data, the rapid expansion of multi-source and multimodal ophthalmic datasets presents unprecedented opportunities. A critical scientific question emerges: how can Artificial Intelligence (AI) unlock the hidden potential embedded within these vast and heterogeneous datasets?

Recently, a comprehensive review titled "Data-driven computational methods in ophthalmology: A multimodal perspective" was published in the international journal Eye Discovery. From a "data-centric" perspective, the article systematically evaluates the scientific value of multimodal ophthalmic data and analyzes cutting-edge advances and future challenges in AI-driven ophthalmic research.

1. Imaging Data: From Macroscopic Scanning to Cellular Visualization

Clinical ophthalmology relies heavily on imaging. AI technology has substantially improved the precision of pathological feature extraction across five core modalities:

Fundus Photography-Based Imaging: As the most widely used modality, AI enables high-precision screening of diabetic retinopathy, glaucoma, and cataracts in color fundus photography. Red-free imaging enhances superficial vascular contrast via AI to facilitate early detection of glaucomatous lesions. Ultra-widefield imaging, combined with AI, significantly improves the capture of peripheral retinal lesions.

Angiographic Imaging: Fluorescein fundus angiography integrated with AI achieves automatic segmentation and quantitative assessment of non-perfusion areas and microaneurysms. Indocyanine green angiography, with AI assistance, enables more precise identification of polypoidal choroidal vasculopathy.

Tomographic and Angiographic Optical Imaging: Optical coherence tomography (OCT) enables automatic quantification of intraretinal fluid; OCT angiography achieves contrast-agent-free capillary-level blood flow visualization; adaptive optics OCT further extends AI into microscopic imaging at the photoreceptor cell level.

Autofluorescence and Laser Scanning Imaging: Fundus autofluorescence, analyzed by AI, enables automated monitoring of retinal pigment epithelium (RPE) metabolism and geographic atrophy assessment. Confocal laser scanning, as a complement to OCT, enhances image sharpness through AI to aid biomechanical analysis.

Anterior Segment Imaging: Slit-lamp photography combined with AI enables automatic cataract grading and corneal inflammation recognition. Ultrasound biomicroscopy, powered by AI, achieves automated measurement of anterior chamber angle structures, supporting glaucoma management.

2. Non-Imaging Data: Decoding the Molecular "Pathogenic Code"

Beyond visual information, AI is advancing omics research to provide molecular support for precision medicine:

Genomics: Identifying risk loci and drug targets associated with numerous eye diseases by processing biobank-scale data through AI.

Transcriptomics: Integrating transcriptome sequencing results with AI to mine expression changes between disease and healthy states, or across specific cell populations.

Proteomics: AI assists in analyzing protein interaction networks relevant to eye diseases and identifying key diagnostic or therapeutic biomarkers.

Metabolomics: Using AI to analyze small-molecule metabolic profiles in blood or aqueous humor, discovering metabolites linked to intraocular pressure regulation.

Electronic Health Records (EHR): Leveraging AI to extract phenotypes from massive clinical records, thereby optimizing disease prediction models.

3. Multimodal Fusion: The "1 + 1 > 2" Diagnostic Power

Ophthalmic diseases involve complex pathophysiological mechanisms that a single data source rarely captures in full. AI is breaking data silos and constructing a comprehensive diagnostic and therapeutic coordinate system through multimodal fusion:

Image-Image Integration: Through techniques such as "knowledge distillation" or "cross-modal generation," AI can "migrate" deep features from one modality (e.g., OCT) to conventional two-dimensional color fundus photography models, enabling routine examinations to capture deep pathological anatomical information.

Non-Imaging Multimodal Data Analysis: By integrating genomics, transcriptomics, and proteomics — combining the molecular "blueprint" and "execution process" — AI enhances diagnostic robustness for hereditary eye diseases and provides stronger biological interpretability.

Image-Non-Image Fusion: This represents the ultimate pathway toward precision ophthalmology. AI achieves cross-modal alignment between intuitive imaging features, deep molecular omics profiles, and unstructured clinical narratives.

4. Scholarly Value and Clinical Trends

This review highlights that although AI has achieved significant success in ophthalmology, future core challenges remain in data standardization and algorithm robustness. The research team proposes that the next breakthrough lies in developing intelligent models with "spatiotemporal response" characteristics and constructing more generalizable, cross-center databases.

This systematic review not only provides clear technical guidance for researchers but also demonstrates the transformative progression of ophthalmic clinical diagnosis — from subjective experience to data-driven decision-making — under digital transformation.

The complete study is accessible via https://doi.org/10.1016/j.edisc.2026.100026

About Eye Discovery

Eye Discovery is an open-access, peer-reviewed international academic journal, with ISSN 3117-4167. It is published quarterly by Elsevier and serves as the official journal of Eye & ENT Hospital of Fudan University, China. 

Eye Discovery  is dedicated to creating a high-end platform for ophthalmologists, scientists, and scholars worldwide to focus on innovative achievements in ophthalmology and interdisciplinary fields, and to promote academic dissemination and exchange.

Website: https://www.sciencedirect.com/journal/eye-discovery
Submit: https://www.editorialmanager.com/edisc/Default.aspx

From 2026 to 2028, the article processing charge ( APC ) will be waived.