Evaluation of a cornea-specialized large language model for diagnostic and management accuracy in complex corneal cases.
Int OphthalmolSep 2, 2026 (epub)
David Mikhail, Daniel Milad, Fares Antaki et al.
✦ AI-curated · Sources linked
In 30 seconds
This randomized controlled trial assessed the effectiveness of a cornea-specialized large language model (LLM) in improving diagnostic and management accuracy in complex corneal cases among three cornea trainees. The study found that diagnostic accuracy improved from 48.7% unaided to 71.8% with the specialized model (p<0.01), while management accuracy showed variable results across clinicians.
Key findings
- Diagnostic accuracy improved from 48.7% unaided to 71.8% with the cornea-specialized model (p<0.01).
- General GPT-4o assistance raised diagnostic accuracy to 69.2% (p<0.04).
- Management accuracy varied, with Ophthalmologist 1 at 82.1% and Ophthalmologist 2 declining to 64.1% (p<0.05).
Why it matters
The study highlights the potential of specialized AI models to enhance diagnostic accuracy in ophthalmology, particularly for clinicians with lower baseline performance. This could lead to improved patient outcomes in complex corneal cases.