Fig. 4
Nomogram performance for predicting axillary lymph node metastasis: Panels (a–c) present the performance of the perioperative model, and panels (d–f) present the performance of the preoperative model. (a) ROC curve of the perioperative nomogram (AUC = 0.917). (b) ROC curves for the nomogram versus individual predictors: quadrant (0.422), tumor size (0.604), molecular subtype (0.490), lympho-vascular invasion (0.700), and axillary sonography (0.853). (c) Calibration curve of the perioperative model showing agreement between predicted and observed probabilities; curves closer to the ideal reference line indicate better calibration. (d) ROC curve of the preoperative nomogram (AUC = 0.898). (e) ROC curves comparing the preoperative nomogram with individual predictors: quadrant (AUC = 0.659), tumor size (AUC = 0.604), molecular subtype (AUC = 0.535), and sonomammography lymph node assessment (AUC = 0.853). (f) Calibration curve of the preoperative model demonstrating good concordance between predicted and observed probabilities.
