Comprehensive Evaluation of Transfer-CNN based Models for Breast Cancer Detection

Abstract

Breast cancer is among the top causes of death for women globally. Detecting this type of cancer at its early stages can reduce the number of early fatalities greatly, due to the fact that manual and biopsy-dependent breast cancer diagnosis takes so long, an automated technique is important for early cancer diagnosis, these Computer-Aided Diagnosis (CAD) approaches are utilized in our study by harnessing deep learning in order to detect breast cancer. In this work, we present a deep analysis of transfer learning based-models   to detect and classify benign from malignant breast tumors. In our study, four well-known CNN architectures are utilized, including ResNet, MobileNet, VGG-16 and EfficientNet, and evaluated using different key performance metrics: accuracy, precision, recall, and F1 score, along with the average training time.