Advancing Breast Cancer Classification in Ultrasound Imaging with ResNetV1 and ResNetV2 for Improved Diagnostic Accuracy

Abstract-Breast cancer, a global health concern with a rising incidence rate, underscores the imperative for improved diagnostic accuracy to enable timely detection and mitigate mortality. Digital mammography, the primary diagnostic tool, faces limitations in detecting tumors within dense breast tissue. In this context, ultrasound imaging emerges as a cost-effective and widely available alternative, particularly for women with dense breast tissue. While ultrasound-based breast imaging holds promise, challenges persist in image interpretation and classification. Machine learning techniques exhibit remarkable potential in enhancing breast cancer diagnosis, and this paper builds upon prior studies to advance the field. Specifically, This research delve into the application of ResNet and ResNetV2 architectures, incorporating transfer learning and fine-tuning. Our findings highlight that fine-tuning significantly boosts model performance, with the ResNet50V2 model achieving an impressive accuracy of 99.11%. This study underscores the relevance of fine-tuning in breast image classification and contributes to ongoing efforts to improve diagnostic accuracy, aiming to reduce delayed diagnoses and decrease breast cancer mortality.