浙江医学2025，Vol.47Issue(17)：1835-1839,后插2,6.DOI:10.12056/j.issn.1006-2785.2025.47.17.2025-269
基于常规超声及剪切波弹性成像构建乳腺癌腋窝淋巴结转移机器学习预测模型
Machine learning models for predicting axillary lymph node metastasis of breast cancer based on conventional ultrasound and shear wave elastography
摘要
Abstract
Objective To construct machine learning models for predicting axillary lymph node metastasis(ALNM)in breast cancer based on conventional ultrasound and shear wave elastography(SWE).Methods A retrospective nested cohort study was adopted by selecting 200 female patients with breast cancer diagnosed by pathology between January 2021 and January 2023 who were admitted to Jinhua Municipal Centeral Hospital were randomly divided into a training set(n=140)and an internal validation set(n=60)in a ratio of 7∶3.They were divided into ALNM positive and negative groups based on pathological results.Preoperative routine ultrasound and SWE parameters were compared between the two groups.Three machine learning methods,namely extreme gradient boosting(XGBoost),decision tree,and logistic regression were used to construct ALNM prediction models.In addition,40 patients with breast cancer diagnosed by pathology in the same hospital were included from June 2023 to December 2024 as an external validation set.ROC curve was used to screen the optimal model and AUC was calculated.Results Among the 200 breast cancer patients,there were 93 ALNM positive patients and 107 negative patients,of which,the training set had 66 positive cases and the internal validation set had 27 cases.In the training set,the maximum diameter of tumor lesion,maximum elastic modulus of lymph nodes(Emax)and standard deviation of elastic modulus(Esd)values in the positive group were significantly higher than those in the negative group,and proportions of tumor stages Ⅲ-Ⅳ,lymph node calcification,blood flow grades Ⅲ-Ⅳ,unclear corticomedullary structure,andⅢ-Ⅳ types in the SWE were also higher(all P＜0.05).ROC showed that the AUC of XGBoost model for predicting ALNM positivity was significantly higher than that of decision tree and logistic regression models in internal and external validation sets(all P＜0.05).Conclusion XGBoost model based on conventional ultrasound and SWE shows the best performance in predicting ALNM positivity of breast cancer.关键词
乳腺癌/腋窝淋巴结转移/机器学习/剪切波弹性成像/弹性模量Key words
Breast cancer/Axillary lymph node metastasis/Machine learning/Shear wave elastography/Elastic modulus引用本文复制引用
朱俊杰,徐琛,周一波..基于常规超声及剪切波弹性成像构建乳腺癌腋窝淋巴结转移机器学习预测模型[J].浙江医学,2025,47(17):1835-1839,后插2,6.基金项目
金华市科技计划项目(2022-4-129) （2022-4-129）
