Mortality Prediction in Patients With Breast Cancer by Artificial Neural Network Model and Elastic Net Regression

Area of Science:

• Oncology
• Biostatistics
• Machine Learning in Healthcare

Background:

• Breast cancer (BC) is a leading cause of mortality in women globally.
• Accurate prediction of BC mortality is crucial for patient management and treatment strategies.
• Identifying predictive models for BC mortality is an ongoing area of research.

Purpose of the Study:

• To evaluate and compare the performance of elastic net regression and artificial neural network (ANN) models in predicting breast cancer mortality.
• To identify key factors influencing breast cancer mortality using these models.
• To assess the diagnostic and prognostic capabilities of machine learning approaches in oncology.

Main Methods:

• A cross-sectional study analyzing data from 2,836 breast cancer patients (2014-2018).
• Utilized elastic net regression and artificial neural network (ANN) models to predict mortality.
• Compared model performance using metrics: sensitivity, specificity, accuracy, AUC, precision, and F1-score.

Main Results:

• Elastic net regression achieved a specificity of 0.814 and accuracy of 0.792.
• Artificial neural network (ANN) demonstrated higher sensitivity (0.66) and AUC (0.704).
• Elastic net regression outperformed ANN in specificity, accuracy, precision, and F1-score.

Conclusions:

• Both elastic net regression and ANN models can be employed to predict breast cancer mortality.
• ANN models offer superior sensitivity and AUC for predicting BC mortality.
• Elastic net regression provides better specificity and accuracy, with morphology, tumor differentiation, and age identified as significant factors affecting mortality.