Time and person sensitive foundation model for disease prediction and risk stratification - PubMed
2 days ago
- #risk stratification
- #medical AI
- #disease prediction
- RETFound Plus is a foundation model (FM) for disease prediction and risk stratification using temporal modeling on fundus photographs.
- Trained on 1,304,292 fundus photographs from 304,345 participants across multiple visits to learn progression-aware representations.
- Improves 5-year risk prediction for systemic diseases (stroke, myocardial infarction, diabetes, hypertension) by +4-10% c-index and ocular diseases (diabetic retinopathy, glaucoma) by +3-7% c-index.
- Enhances risk stratification for systemic diseases with 1.2-2.1-fold higher hazard-ratio trend.
- Validated across multi-regional, multi-ethnic datasets from the UK, US, Singapore, Hong Kong, and Denmark.
- Authors have various competing interests, including consultancy roles and equity in medical technology companies.
- Funded by multiple grants including Wellcome Award, National Science and Technology Major Project, and National Key Research and Development Program of China.