@article{GS156004,
author = {Lifen Cai and Ying Zhao and Jie Wang and Yilei Ruan},
title = {Integrating intratumoral and peritumoral ultrasound radiomics with clinicopathological features to predict axillary lymph node metastasis in early breast cancer},
journal = {Gland Surgery},
volume = {15},
number = {6},
year = {2026},
keywords = {},
abstract = {Background: XXX. This study aimed to develop and validate a predictive model for axillary lymph node metastasis (ALNM) in early breast cancer (EBC) based on preoperative ultrasound-derived intratumoral and peritumoral radiomics features integrated with clinicopathological features.Methods: In this retrospective study, 327 patients with EBC were enrolled. Clinicopathological variables and preoperative ultrasound images were collected, and patients were randomly assigned to a training cohort (n=228) and a validation cohort (n=99) at a 7:3 ratio. Intratumoral, 4-mm peritumoral, and combined intratumoral-peritumoral regions of interest were delineated using ITK-SNAP software. Logistic regression (LR), support vector machine, and random forest machine learning (ML) algorithms were applied to construct radiomics models based on intratumoral, peritumoral, and combined regions, respectively. A clinical-radiomics combined model was subsequently developed by integrating independent clinical predictors. Model performance was evaluated using receiver operating characteristic (ROC) curves, calibration curves, decision curve analysis (DCA), and Shapley additive explanations (SHAP).Results: Multivariable LR analysis identified maximal diameter and lymphovascular invasion as independent predictors of ALNM. Among the evaluated ML algorithms, the LR-based combined intratumoral and peritumoral radiomics model demonstrated the best performance, achieving an area under the curve (AUC) of 0.868 [95% confidence interval (CI): 0.795–0.933] in the validation cohort. Incorporation of independent clinical predictors further improved model performance, with the clinical-radiomics combined model yielding a validation AUC of 0.909 (95% CI: 0.851–0.960). Calibration curves and DCA indicated good agreement between predicted and observed outcomes and demonstrated favorable clinical utility. SHAP analysis revealed that radiomics features contributed predominantly to model predictions.Conclusions: A predictive model integrating intratumoral and peritumoral ultrasound radiomics features with clinicopathological variables enables accurate preoperative identification of ALNM in patients with EBC. This model may serve as a reliable and non-invasive tool to guide individualized axillary management strategies.},
issn = {2227-8575}, url = {https://gs.amegroups.org/article/view/156004}
}