Related Experiment Video
Updated: Sep 19, 2026
Comparison of Predictive Performance of Three Lymph Node Staging Systems in Colorectal Signet Ring Cell Carcinoma Based on Machine Learning Model
Published on: April 18, 2025
Predictive modeling of axillary web syndrome in Chinese postoperative breast cancer patients using interpretable
Jiali Du1, Jing Yang1, Xujin Chen1
1Department of Breast Surgery, Sichuan Clinical Research Center for Cancer, Sichuan Cancer Hospital &Institute, Sichuan Cancer Center, Affiliated Cancer Hospital of University of Electronic Science and Technology of China, Chengdu, China.
Objective:
This study aimed to develop and validate a series of risk prediction models for axillary web syndrome (AWS) after breast cancer surgery using machine learning algorithms, enabling postoperative risk identification to facilitate timely targeted interventions.
Methods:
This retrospective cross-sectional study was conducted at a tertiary cancer hospital between April 2021 and July 2024. After collecting clinical data, the least absolute shrinkage and selection operator (LASSO) regression was employed to screen variables with predictive value for AWS. The dataset was randomly divided into a development set and a test set at a 7:3 ratio. Nine classification models were constructed, and their performance was evaluated based on accuracy, precision, recall, F1-score, and area under the curve (AUC). The optimal model was selected according to the AUC value. Model interpretation was performed using SHAP analysis and feature importance ranking.
Results:
A total of 655 patients were included, with 459 in the development set and 196 in the test set. The AWS incidence rate was 26.9% (176 cases). LASSO regression identified 14 most predictive features, and SHAP value analysis revealed that lymph node surgical approach and lymphedema were the most critical variables. The logistic regression (LR) model demonstrated the best overall predictive performance (AUC = 0.72).
Conclusion:
This study confirmed that the LR model exhibited superior accuracy and AUC value combinations, effectively identifying high-risk AWS patients. The findings provide a basis for clinical practitioners to identify high-risk patients and implement targeted interventions.