AUTHOR=Peng Xuemei , Ai Yanfang , Xu Weiying , Hong Jingling , Li Qinyan , Liu Jianfen 

TITLE=Machine learning-based personalized risk prediction model for breast cancer-related lymphedema after surgery

JOURNAL=Frontiers in Oncology

VOLUME=Volume 15 - 2025

YEAR=2025

URL=https://www.frontiersin.org/journals/oncology/articles/10.3389/fonc.2025.1729340

DOI=10.3389/fonc.2025.1729340

ISSN=2234-943X

ABSTRACT=ObjectiveBreast cancer-related lymphedema (BCRL) is a common postoperative complication that significantly impairs patients’ quality of life. This study aims to develop a machine learning-based personalized risk prediction model for BCRL by integrating multimodal clinical and behavioral data, thereby providing scientific support for early identification and intervention in high-risk individuals.MethodsClinical and follow-up data were collected from patients who underwent breast cancer surgery between June 2020 and June 2025. A total of 38 variables were analyzed using the Least Absolute Shrinkage and Selection Operator (LASSO) method for feature selection. Nine machine learning algorithms were developed, and their performance was evaluated using metrics including accuracy, sensitivity, specificity, positive predictive value, negative predictive value, F1-score, and the area under the receiver operating characteristic curve (AUC). The optimal model was further interpreted using Shapley Additive Explanations (SHAP) to enable individualized risk assessment.ResultsA total of 368 eligible patients were included and randomly divided into a training set (n = 257) and a validation set (n = 111) at a 7:3 ratio. Among them, 98 patients (26.63%) developed BCRL. LASSO regression identified 12 features most predictive of BCRL. Among all models, logistic regression demonstrated the best performance, with an AUC of 0.937, accuracy of 0.793, sensitivity of 0.937, and specificity of 0.740 in the validation set. BMI, lymph node dissection level, and lymph node status were identified as the most influential predictors contributing to model performance.ConclusionA logistic regression model, combined with SHAP-based interpretation, enables personalized risk prediction of BCRL in postoperative breast cancer patients. This approach may provide robust support for clinical risk stratification and intervention planning.