Related Experiment Videos
Development and validation of an interpretable machine learning model for predicting 5-year recurrence in breast
Shaoda Meng1, Sicheng Liu2, Minghua Lai1
1Department of Breast and Thyroid Surgery, The First People's Hospital of Yunnan Province, Kunming, Yunnan, China.
Frontiers in Medicine
|June 26, 2026
Summary
This study developed an interpretable machine learning model to predict breast cancer recurrence, outperforming traditional staging. The model refines risk stratification for personalized treatment decisions.
Area of Science:
- Oncology
- Biostatistics
- Machine Learning in Medicine
Background:
- Breast cancer recurrence prediction is crucial for adjuvant therapy.
- The TNM staging system has limitations due to biological heterogeneity.
- A need exists for more precise individual risk stratification.
Purpose of the Study:
- To develop an interpretable machine learning model for breast cancer recurrence risk stratification.
- To refine prognostic accuracy beyond traditional staging methods.
- To aid in personalized clinical decision-making for adjuvant therapy.
Main Methods:
- Retrospective analysis of 578 breast cancer patients.
- Development of an Extreme Gradient Boosting (XGBoost) model using 15 clinicopathological features.
- Application of SHapley Additive exPlanations (SHAP) for model interpretability and construction of a visual nomogram.
Main Results:
- The XGBoost model achieved an AUC of 0.877 in the validation cohort, outperforming logistic regression and TNM staging.
- SHAP analysis highlighted Ki-67 index and lymph node status as key predictors.
- The model effectively stratified patients within TNM Stages II and III into distinct risk groups.
Conclusions:
- The XGBoost-based framework offers a robust and interpretable tool for predicting 5-year breast cancer recurrence.
- This approach provides superior prognostic accuracy compared to standard anatomical staging.
- The model shows promise for facilitating personalized treatment strategies.
