Machine Learning Model for Predicting Axillary Lymph Node Metastasis in Clinically Node Positive Breast Cancer Based...

Area of Science:

• Oncology
• Medical Imaging
• Machine Learning

Background:

• Accurate prediction of axillary lymph node metastasis is crucial for staging and treatment of early-stage breast cancer.
• Current methods may not fully capture all relevant prognostic information.

Purpose of the Study:

• To develop and validate a machine learning model integrating clinical features and ultrasound radiomics for predicting axillary lymph node metastasis.
• To compare the diagnostic performance of different machine learning models and identify the optimal algorithm.

Main Methods:

• Retrospective analysis of clinical, ultrasound, and pathological data from 321 early-stage breast cancer patients.
• Construction of a clinical feature model and a radiomics feature model.
• Development and comparison of six machine learning models, including XGBoost, for a joint prediction model.
• Utilization of Shapley Additive Explanations (SHAP) for model interpretability.

Main Results:

• The joint prediction model using XGBoost achieved high diagnostic performance with AUCs of 0.917 (training) and 0.905 (validation).
• The radiomics model, incorporating intratumoral and peritumoral features, outperformed the clinical model.
• SHAP analysis highlighted the significant contribution of radiomic features to the model's predictive power.

Conclusions:

• An integrated XGBoost model combining clinical and radiomic features shows significant value in predicting axillary lymph node metastasis.
• This model aids in preoperative surgical strategy selection and prognosis evaluation for breast cancer patients.
• Enhanced model interpretability through SHAP facilitates clinical application and trust.