Breast cancer recurrence prediction with ensemble methods and cost-sensitive learning

Area of Science:

• Oncology
• Biostatistics
• Machine Learning

Background:

• Breast cancer is a leading cause of cancer death in women worldwide.
• While treatments improve remission rates, cancer recurrence remains a significant threat, necessitating accurate prediction for improved outcomes.
• Early prediction of breast cancer recurrence is critical for timely intervention and reducing mortality.

Purpose of the Study:

• To develop and evaluate a predictive model for breast cancer recurrence using clinical data.
• To identify key clinical features associated with breast cancer recurrence.
• To compare the performance of AdaBoost and cost-sensitive learning classifiers for recurrence prediction.

Main Methods:

• Data preprocessing and feature selection were performed on 1,061 patient records, identifying six key features.
• Resampling techniques were employed to address class imbalance in the dataset.
• Two classifiers, AdaBoost and cost-sensitive learning, were constructed and evaluated using three-fold cross-validation.

Main Results:

• The AdaBoost classifier achieved an accuracy of 0.973 and a sensitivity of 0.675.
• The combined AdaBoost and cost-sensitive learning model yielded an accuracy of 0.468 and a high sensitivity of 0.947.
• The combined model demonstrated a strong ability to minimize false dismissals, crucial for early detection.

Conclusions:

• The developed prediction model, particularly the combination of AdaBoost and cost-sensitive learning, shows promise as a decision support tool.
• This model can aid in the evaluation of follow-up visits, facilitating early intervention and advanced treatments.
• Effective prediction of breast cancer recurrence can significantly contribute to lowering cancer mortality rates.