Skip to main content
. 2025 Oct 13;12:1655302. doi: 10.3389/fmed.2025.1655302

FIGURE 6.

Two panels show ROC curves comparing clinical, image, and combined data for train (A) and test (B) sets. Panel A: Train curves with areas under the curve (AUC) are 0.676 for clinical, 0.661 for image, and 0.970 for combined data. Panel B: Test curves with AUCs are 0.644 for clinical, 0.579 for image, and 0.908 for combined data. The curves are plotted against a diagonal reference line.

ROC curves of different models in training and test cohorts. (A) Training cohort ROC curves the combined model demonstrates the best performance, with an AUC of 0.970 (95% confidence interval: 0.937–0.997), indicating high accuracy in distinguishing between good and poor scar prognoses. The clinical model has an AUC of 0.676 (95% CI: 0.545–0.790), and the image model has an AUC of 0.661 (95% CI: 0.519–0.802). The AUC values are accompanied by their respective 95% confidence intervals, providing an indication of the reliability of the model performance. (B) Test cohort ROC curves the combined model still shows excellent performance, with an AUC of 0.908 (95% CI: 0.783–1.000). The clinical model has an AUC of 0.644 (95% CI: 0.400–0.848), and the image model has an AUC of 0.579 (95% CI: 0.356–0.827). The performance of the combined model in the test cohort indicates its good generalization ability, while the clinical and image models show relatively lower and less stable performance.