Deep Learning-Based Automated Breast Cancer Ultrasound Image Classification: A Study
R. Sivakumar
Abstract
Breast cancer is a major public health concern worldwide, and early detection has the potential to improve treatment outcomes. This study evaluates various DL models for their effectiveness in identifying breast cancer using ultrasound images. Data preprocessing, Feature extraction, and binary classification have been thoroughly conducted throughout the study procedure. MobileNet and MobileNetV2 offer computational efficiency, whilst the Xception model excels in feature extraction. The intricate design of VGG16 and VGG19 ensures consistent accuracy and comprehensive feature representation. Among them, VGG19 is the model that has the best categorisation accuracy, exceeding $96 \%$. The performance statistic encompasses accuracy, precision, recall, and F1-score, utilising five-fold cross-validation. This work illustrates the ability of DL to improve clinical decision-making in diagnostic imaging and facilitate early diagnosis of breast cancer.