AI Is Learning to Spot Multiple Eye Diseases at Once—Here's Why That Matters for Your Vision

A new artificial intelligence system called RetExpert can detect up to 39 different eye diseases from a single retinal photograph, addressing a major gap in how ophthalmologists currently screen for vision-threatening conditions. Most existing AI tools struggle when patients have multiple eye diseases at the same time—a common real-world scenario that can confuse diagnostic systems. RetExpert was designed specifically to handle these complex cases, making it potentially transformative for eye disease screening in clinical settings.

Why Can't Current AI Systems Handle Multiple Eye Diseases?

Fundus diseases—conditions affecting the retina and back of the eye—are among the leading causes of vision loss globally. The challenge for artificial intelligence isn't just detecting one disease; it's accurately identifying multiple conditions that often occur together. Researchers found that conventional AI models struggle with imbalanced data, uncertainty when multiple diseases are present, and confusion between similar-looking conditions.

When patients have diabetes, for example, they might simultaneously develop diabetic retinopathy (damage from high blood sugar) and age-related macular degeneration (AMD), a separate age-related condition. Traditional AI systems often misidentify one disease as another or miss conditions entirely when they coexist. This diagnostic confusion can delay treatment and worsen outcomes.

How Does RetExpert Actually Work?

RetExpert incorporates several innovative features that make it more reliable than previous systems. The framework uses what researchers call "adaptive knowledge units" combined with a stochastic one-hot activation module to improve how well the system generalizes across different patient populations and imaging equipment. It also integrates a fundus disease co-occurrence matrix—essentially a medical knowledge map showing which diseases commonly appear together—to reduce diagnostic confusion.

The system was tested on 15 different public and private datasets, demonstrating strong performance across diverse populations and clinical settings. This extensive validation is crucial because AI systems trained on one population sometimes fail when applied to different demographic groups or healthcare environments.

Steps to Understanding How AI Screening Could Change Your Eye Care

• Faster Detection: Instead of waiting for an ophthalmologist to manually examine retinal images, AI can screen multiple conditions simultaneously from a single photograph, potentially identifying diseases earlier when treatment is most effective.
• Reduced Diagnostic Errors: By accounting for disease co-occurrence patterns, RetExpert minimizes the confusion between similar-looking conditions that can lead to missed diagnoses or incorrect treatment plans.
• Broader Access to Screening: AI-powered systems could enable primary care clinics and community health centers to perform preliminary eye disease screening, expanding access beyond specialized ophthalmology practices.
• Real-World Adaptability: The system uses test-time adaptation—a technique that allows it to adjust to new clinical environments without requiring complete retraining, making deployment more practical and cost-effective.

What Eye Diseases Can RetExpert Detect?

The system can identify 39 different fundus diseases and conditions, including some of the most common vision threats. These range from diabetic retinopathy (affecting people with diabetes) to age-related macular degeneration, glaucoma-related changes, retinal vein occlusions, and inherited retinal diseases. The ability to detect this many conditions simultaneously represents a significant leap forward, since most current screening tools focus on one or two specific diseases.

For patients with multiple risk factors—such as older adults with both diabetes and a family history of glaucoma—this comprehensive approach could catch several threats in a single screening visit, potentially preventing vision loss that might otherwise go undetected.

When Might This Technology Reach Your Eye Doctor?

While RetExpert shows promise in research settings, the path to clinical use involves several steps. The system needs validation in real-world clinical environments, integration with existing electronic health record systems, and regulatory approval in various countries. However, the research demonstrates that AI-powered multi-disease detection is technically feasible and clinically viable, suggesting that widespread adoption could happen within the next few years.

The timing is particularly important because retinal diseases remain leading causes of preventable blindness worldwide. Early detection through improved screening tools could significantly reduce vision loss, especially in regions with limited access to specialized eye care.

What Does This Mean for People at Risk for Eye Disease?

For anyone with diabetes, a family history of glaucoma, or age-related vision concerns, improved AI screening could mean earlier intervention and better outcomes. Rather than waiting for symptoms to appear—which often happens too late for some eye diseases—AI-powered screening could identify threats during routine eye exams when treatment is most effective. This is particularly valuable for conditions like glaucoma, which often progresses silently without noticeable symptoms until significant vision loss has already occurred.

The development of RetExpert also signals a broader shift in ophthalmology toward precision medicine, where screening and treatment are tailored to individual disease patterns rather than treating each condition in isolation.