Can machine learning models predict if my early Age-Related Macular Degeneration will get worse?

Age-related macular degeneration (AMD) is a leading cause of vision loss. Early AMD may stay stable or progress to advanced stages like geographic atrophy or neovascular AMD. Machine learning models analyze patterns in eye scans and health records to estimate your personal risk of worsening. While these tools are not yet standard in clinics, research shows they can predict progression with reasonable accuracy.

What the research says

A 2025 study developed machine learning models using multimodal fundus images from 324 AMD patients to predict progression from early to late AMD over 3 years. The best model (XGBoost) achieved an AUC of 0.895, meaning it correctly distinguished progressors from non-progressors about 90% of the time. Key predictors included drusen area and pigmentary abnormalities ⁶. Another study used health records to predict early-onset AMD (diagnosis before age 70) with gradient-boosted decision trees, achieving AUCs of 0.74-0.80 ⁸. For patients already on treatment, machine learning models using OCT scans and clinical data can predict visual outcomes at 9 months for anti-VEGF drugs like ranibizumab ⁹ and faricimab ¹⁰. These models combine imaging features (retinal fluid, layer thickness) with patient characteristics. However, most models are still in research stages and need validation in diverse populations before routine clinical use.

What to ask your doctor

• Are there any ongoing studies or clinical trials at this center using machine learning to predict AMD progression?
• What are the main risk factors for my AMD worsening, such as drusen size, pigment changes, or smoking?
• How often should I have follow-up eye exams and imaging to monitor for progression?
• Could my other health conditions, like sleep apnea or high CRP levels, affect my AMD risk?
• What lifestyle changes (diet, supplements, quitting smoking) might lower my chance of progression?