Breast Cancer Prognosis Using a Machine Learning Approach

Area of Science:

• Oncology
• Biomedical Informatics
• Machine Learning

Background:

• Machine learning (ML) is emerging for cancer prognosis.
• Predicting individual cancer patient outcomes is crucial for treatment.

Purpose of the Study:

• To evaluate an ML-based decision support system (DSS) combined with random optimization (RO).
• To extract prognostic information from routinely collected breast cancer patient data.

Main Methods:

• Developed a DSS model using training data (n=318).
• Applied random optimization to extract prognostic features.
• Evaluated model performance on a testing set (n=136).

Main Results:

• Achieved a C-index of 0.84 for progression-free survival.
• Demonstrated 86% accuracy in patient outcome prediction.
• Successfully stratified patients into low- and high-risk groups (HR=10.9, p<0.0001).

Conclusions:

• ML algorithms and RO models integrated with EHR data can yield valuable prognostic information.
• This approach has the potential to advance personalized cancer medicine.
• Further validation in multicenter prospective studies is recommended.