Developing a Novel Machine Learning-Based Classification Scheme for Predicting SPCs in Breast Cancer Survivors.

Area of Science:

• Oncology
• Machine Learning
• Biostatistics

Background:

• Increased diagnosis of second primary cancers (SPCs) in breast cancer survivors due to effective screening and therapies.
• Need for predictive models to identify high-risk individuals for SPCs.

Purpose of the Study:

• To develop a novel machine learning classification scheme for predicting SPC risk in breast cancer survivors.
• To identify key risk factors associated with SPC development.

Main Methods:

• Utilized the XGBoost classifier with strategies including transformation, resampling, clustering, and ensemble learning.
• Evaluated model performance based on balanced accuracy.
• Identified important risk factors through experimental analysis.

Main Results:

• The XGBoost classifier combined with resampling and clustering strategies demonstrated the best prediction accuracy.
• Identified age, radiotherapy/surgery sequence, surgical margins, human epidermal growth factor, high-dose clinical target volume, and estrogen receptors as significant risk factors for SPCs.
• These factors are crucial for early detection of SPCs in breast cancer patients.

Conclusions:

• The proposed machine learning scheme effectively supports the identification of influential factors in the treatment trajectory.
• Adaptive machine learning models incorporating significant variables are crucial for optimal SPC risk prediction in breast cancer survivors.