Application of SHAP for Explainable Machine Learning on Age-Based Subgrouping Mammography Questionnaire Data for...

Area of Science:

• Oncology
• Biostatistics
• Computational Biology

Background:

• Mammography is the primary tool for breast cancer screening.
• Understanding breast cancer risk factors is crucial but debated.
• Machine learning (ML) and SHAP offer advanced methods for risk factor analysis.

Purpose of the Study:

• To utilize ML and SHAP to analyze and rank breast cancer risk factors.
• To compare the impact of risk factors across different age groups.
• To evaluate factors influencing positive mammography outcomes.

Main Methods:

• Data from women in a breast cancer screening program (2017-2021) were used.
• Three ML models (lasso, XGBoost, RF) were applied, with Random Forest showing superior performance.
• SHAP values were used with the RF model to interpret risk factor significance.

Main Results:

• The top five risk factors identified were age at menarche, education level, parity, breast self-examination, and BMI.
• Differences in reproductive lifespan and BMI impact varied between younger and older age groups.
• SHAP analysis provided individualized risk factor rankings.

Conclusions:

• ML and SHAP effectively identify and rank significant breast cancer risk factors.
• Individualized risk factor profiles can be generated for targeted screening and prevention.
• This approach supports advancements in personalized medicine for breast cancer.