Machine Learning Models Utilizing Oxidative Stress Biomarkers for Breast Cancer Prediction: Efficacy and Limitations...

Area of Science:

• Biomedical engineering
• Machine learning in oncology
• Oxidative stress research

Background:

• Breast cancer and sentinel lymph node (SLN) metastasis pose significant public health challenges.
• Accurate classification and metastasis detection are crucial for patient outcomes.
• Oxidative stress biomarkers offer potential for novel diagnostic approaches.

Purpose of the Study:

• To apply a Random Forest machine learning model using oxidative stress biomarkers for breast cancer classification.
• To assess the model's efficacy in detecting sentinel lymph node (SLN) metastasis.
• To evaluate the potential of machine learning and biomarkers in improving breast cancer diagnostics.

Main Methods:

• Utilized a Random Forest model with leave-one-out validation.
• Employed oxidative stress biomarkers including lipid peroxidation, antioxidant capacity, superoxide dismutase, catalase, and glutathione peroxidase.
• Applied SMOTE technique for class balancing in the metastasis dataset.

Main Results:

• Achieved high accuracy (0.996) in classifying breast cancer status.
• Demonstrated more limited performance in detecting SLN metastases (accuracy = 0.854).
• Results are from a retrospective cohort and require prospective validation for comparison with mammography screening.

Conclusions:

• The Random Forest model shows robust performance in distinguishing breast cancer from healthy individuals.
• Significant challenges remain in accurately predicting metastatic disease using this model.
• Future research should incorporate additional biomarkers, longitudinal data, and explainable AI for enhanced metastasis prediction and precision medicine.