Detection of Benign/Malignent Breast Cancer with ResNet Models

The incidence of breast cancer (BC) in women is increasing due to many factors, contributing significantly to the cancer burden in countries. The clinical identification of breast cancer is conducted utilising pictures from a selected modality, with Breast Ultrasound Imaging (BI) being a prevalent and regularly utilised imaging technique due to its safety and cost-effectiveness. The outcome of BC identification in BI can inform the planning and implementation of requisite treatment to address the disease. This study proposes a Deep Learning (DL) method utilising ResNet (RN) variants, with the results for the SoftMax classifier presented and reviewed. The phases in the created system encompass image acquisition and resizing, feature extraction utilising the RN model, and classification together with performance assessment by 3-fold cross-validation. The experimental results of this work demonstrate that the RN-variants based detection of BC attains accuracy over 99%.