An eyecare foundation model for clinical assistance
13 days ago
- #Clinical trials
- #AI in healthcare
- #Ophthalmology
- Developed EyeFM, a multimodal vision–language eyecare copilot, for clinical assessments of foundation models.
- Pretrained on 14.5 million ocular images from five imaging modalities paired with clinical texts from global, multiethnic datasets.
- Conducted a multifaceted evaluation including retrospective validations, multicountry efficacy validation, and a double-masked randomized controlled trial (RCT).
- The RCT involved 668 participants in China, showing higher correct diagnostic rate (92.2% vs. 75.4%) and referral rate (92.2% vs. 80.5%) with EyeFM.
- Improved standardization score of clinical reports (median 33 vs. 37) and higher compliance with self-management (70.1% vs. 49.1%) and referral suggestions (33.7% vs. 20.2%).
- Post-deployment evaluations indicated strong user acceptance, demonstrating EyeFM's potential to improve ophthalmologists' performance and patient outcomes.