
In a recent study evaluating the predictive capabilities of the artificial intelligence (AI) model ChatGPT in diabetic retinopathy (DR) risk assessment, findings indicate a high level of consistency. The study utilized a dataset of 111 diabetic patients and compared ChatGPT responses to clinical DR diagnoses. Results revealed excellent reliability (ICC: 0.936 with CST, 0.915 without CST). Sensitivity was 67% (with CST) and 73% (without CST), while specificity was 68% (with CST) and 54% (without CST). Despite fair agreement (Cohen's kappa), the study suggests ChatGPT's potential as a preliminary DR screening tool, requiring further optimization for clinical use.
Like
Save
Share