| Title: | AI-Driven Multimodal Prediction of 10-Year Incident Glaucoma Integrating Genetic and Deep Learning-Derived Imaging Features |
| Journal: | Ophthalmology Science |
| Published: | 1 Jun 2026 |
| DOI: | https://doi.org/10.1016/j.xops.2026.101292 |
| URL: | https://www.ophthalmologyscience.org/article/S2666-9145(26)00230-7/pdf |
Publication 20734
WARNING: the interactive features of this website use CSS3, which your browser does not support. To use the full features of this website, please update your browser.
Abstract
Purpose Glaucoma is the leading cause of irreversible blindness worldwide. It often remains asymptomatic until advanced stages. Hence, accurate prediction of glaucoma is crucial for timely intervention to prevent vision loss. Design Population-based prospective cohort study. Subjects, Participants, and/or Controls UK Biobank participants who had available color fundus photographs (CFP), genetic data, and ocular measurements, and were free of glaucoma at baseline. The primary analytic cohort for strictly defined incident POAG comprised 340 cases and 9,374 controls; the secondary broadly defined POAG cohort comprised 1,241 cases and 34,216 controls. Methods We developed an interpretable, multimodal machine learning (ML) framework to predict 10-year incident POAG. The framework integrated five feature domains: deep learning (DL)-derived CFP features, ocular measurements, polygenic risk scores (PRS), physical and lifestyle factors, and electronic health records. DL-derived imaging features included a CFP glaucoma score and automated vertical cup-to-disc ratio estimation. We benchmarked ten machine learning algorithms to identify the optimal model. We applied SHapley Additive exPlanations (SHAP) for model interpretability and feature contribution. Main Outcome Measures The primary outcome was incident POAG defined by ICD-10 code H40.1. A secondary broad POAG phenotype incorporated H40.1, H40.0, and H40.9, while excluding other glaucoma subtypes. Results In the strictly defined POAG analysis, XGBoost achieved the strongest overall performance for 10-year incident glaucoma prediction, with an area under the receiver operating characteristic curve (AUC) of 0.927 (95% CI, 0.895-0.959), high sensitivity (0.685) at a fixed specificity of 0.95, and excellent calibration (Brier score, 0.024). The corresponding XGBoost model for broadly defined POAG showed modestly lower discrimination while maintaining similarly strong calibration. Incremental modeling demonstrated that both PRS and DL-derived imaging features provided complementary predictive value beyond traditional clinical factors. SHAP analysis identified the CFP DL score, age, PRS, and intraocular pressure as the most influential predictors. A reduced model using only the top four SHAP-ranked features retained performance comparable to the all-features multimodal model. Conclusions Our multimodal ML framework integrating genetic and DL-derived imaging features enables accurate and interpretable prediction of incident POAG.</p>
2 Authors
- Fengze Wu
- Xiaoyi Raymond Gao
1 Application
| Application ID | Title |
|---|---|
| 23424 | Ocular Genetics in the UK Biobank |
Enabling scientific discoveries that improve human health