Search

Objective: To determine the appropriateness of ophthalmology recommendations from an online chat-based artificial intelligence model to ophthalmology questions. Patients and Methods: Cross-sectional qualitative study from April 1, 2023, to April 30, 2023. A total of 192 questions were generated spanning all ophthalmic subspecialties. Each question was posed to a large language model (LLM) 3 times. The responses were graded by appropriate subspecialists as appropriate, inappropriate, or unreliable in 2 grading contexts. The first grading context was if the information was presented on a patient information site. The second was an LLM-generated draft response to patient queries sent by the electronic medical record (EMR). Appropriate was defined as accurate and specific enough to serve as a surrogate for physician-approved information. Main outcome measure was percentage of appropriate responses per subspecialty. Results: For patient information site-related questions, the LLM provided an overall average of 79% appropriate responses. Variable rates of average appropriateness were observed across ophthalmic subspecialties for patient information site information ranging from 56% to 100%: cataract or refractive (92%), cornea (56%), glaucoma (72%), neuro-ophthalmology (67%), oculoplastic or orbital surgery (80%), ocular oncology (100%), pediatrics (89%), vitreoretinal diseases (86%), and uveitis (65%). For draft responses to patient questions via EMR, the LLM provided an overall average of 74% appropriate responses and varied by subspecialty: cataract or refractive (85%), cornea (54%), glaucoma (77%), neuro-ophthalmology (63%), oculoplastic or orbital surgery (62%), ocular oncology (90%), pediatrics (94%), vitreoretinal diseases (88%), and uveitis (55%). Stratifying grades across health information categories (disease and condition, risk and prevention, surgery-related, and treatment and management) showed notable but insignificant variations, with disease and condition often rated highest (72% and 69%) for appropriateness and surgery-related (55% and 51%) lowest, in both contexts. Conclusion: This LLM reported mostly appropriate responses across multiple ophthalmology subspecialties in the context of both patient information sites and EMR-related responses to patient questions. Current LLM offerings require optimization and improvement before widespread clinical use.

Purpose: To develop a model to predict corneal improvement after Descemet membrane endothelial keratoplasty (DMEK) for Fuchs endothelial corneal dystrophy (FECD) from Scheimpflug tomography. Design: Cross-sectional study. Participants: Forty-eight eyes (derivation group) and 45 eyes (validation group) with a range of severity of FECD undergoing DMEK. Methods: Scheimpflug images were obtained before and after DMEK. Before DMEK, pachymetry map and posterior elevation map patterns were quantified by a special image analysis program measuring tomographic features of edema (loss of regular isopachs, displacement of the thinnest point of the cornea, posterior surface depression). Image-derived novel parameters were combined with instrument-derived parameters, and the relative influences of parameters associated with the change in central corneal thickness (CCT) after DMEK in the derivation group were determined by using a gradient boosting machine learning model. The parameters with highest relative influence were then fit in a linear regression model. The derived model was applied to the validation group. Correlations and agreement were assessed between predicted and observed changes in CCT. Main Outcome Measures: Predictive power (R2) and mean difference between predicted and observed change in CCT. Results: The gradient boosting machine model identified 4 novel parameters of isopach circularity and eccentricity and 1 instrument-derived parameter (posterior surface radius); preoperative CCT was a poor predictor. In the derivation group, the model strongly predicted the change in CCT after DMEK (R2 = 0.80; 95% confidence interval [CI], 0.71–0.89) and the mean difference between predicted and observed change was, by definition, 0 μm. When the same 5 parameters were fit to the validation group, the model performed very highly (R2 = 0.89; 95% CI, 0.84–0.94). When the coefficient estimates from the derivation model were used to predict the change in CCT in the validation group, the predictive power was also high (R2 = 0.78; 95% CI, 0.68–0.88), and the mean difference was 4 μm (predicted minus observed). Conclusions: Scheimpflug tomography maps of corneas with FECD can predict the improvement in CCT after DMEK, independent of preoperative corneal thickness measurement. The model could be applied in clinical practice or for clinical research of FECD.