The Impact of Biomarker Combination on the Breast Cancer Detection Using the Coimbra Dataset and Neural Networks

Abstract

Breast cancer (BC) is the most prevalent type of cancer in women around the globe. Cancer diagnoses have recently included tissue-based biomarkers, protein-based biomarkers, and molecular-based biomarkers. Several machine learning methods have been created and applied to successfully use biomarkers in breast cancer diagnosis and detection. In this study, we investigate the impact of particular biomarker combinations on the final accuracy and performance using the method of normalized features and neural networks. This article provides new findings on how biomarker combination affects cancer detection effectiveness using normalized features and a FNN model. The research found overall performance of detection with Insulin does not show any additional or distinct difference in accuracy, highlighting Insulin's modest role in identifying BC cases compared to Glucose, Resistin, and HOMA with classification accuracy between 83% and 88%.Moreover, HOMA, Glucose, BMI, and Leptin play an essential part in identifying BC when their normalized values are compared to the average of healthy and BC samples with an accuracy of 95% and sensitivity of 92%.