The aim is to evaluate the accuracy of an artificial intelligence (AI) model, developed using ChatGPT 4.0, in predicting short-term NIHSS (National Institutes of Health Stroke Scale) scores in acute ischemic stroke patients. A retrospective study was conducted on 230 patients with confirmed ischemic stroke undergoing reperfusion therapy. The dataset included demographic, clinical parameters (including baseline NIHSS and mRS), treatment times, and 24-hour post-treatment brain imaging results. An AI model was developed using ChatGPT 4.0 to predict 7th-day NIHSS scores. Model predictions were compared with actual NIHSS scores to assess predictive accuracy using statistical methods, including Pearson correlation and regression analysis. The mean age of 230 patients was 68.41±15.07 years. The AI model demonstrated positive correlation with actual NIHSS scores (r = 0.513, p < 0.001). The model explained 26.3% of the variance in NIHSS scores (R² = 0.263). In contrast, predictions without the model integration showed a weak correlation with actual scores (r = 0.054, p = 0.500), suggesting limited effectiveness. The AI model shows potential for predicting short-term stroke outcomes. However, further refinement is needed to improve accuracy, especially by controlling for patient heterogeneity and enhancing data completeness.
Acır et al. (Fri,) studied this question.