Machine learning approaches for predicting 5-year breast cancer survival: A multicenter study

Area of Science:

• Oncology
• Medical Informatics
• Machine Learning in Healthcare

Background:

• Breast cancer remains a significant global health challenge.
• Accurate survival prediction is crucial for personalized treatment strategies.
• Identifying prognostic factors aids in better patient management.

Purpose of the Study:

• To develop and validate machine learning models for predicting breast cancer survival.
• To identify key prognostic factors influencing survival in Taiwanese women.
• To provide a clinical decision support tool for breast cancer treatment.

Main Methods:

• Retrospective study using electronic medical records from Taiwan (2009-2020).
• Inclusion of 3914 female patients diagnosed with primary breast cancer.
• Development and evaluation of nine machine learning algorithms, including artificial neural networks (ANNs).

Main Results:

• The ANN model achieved the highest Area Under the Curve (AUC) of 0.95.
• Key predictors identified: cancer stage, tumor size, age at diagnosis, surgery, and body mass index.
• Models demonstrated high accuracy (0.90) and negative predictive value (0.94).

Conclusions:

• Accurate 5-year breast cancer survival predictive models were successfully established.
• The study identified critical factors affecting survival in Taiwanese women.
• Findings can inform clinical practice and treatment decisions for breast cancer.