Deep/Federated Learning Algorithms for Ultrasound Breast Cancer Image Enhancement
Sarah M. Waly, Radwa Taha, Mohamed A. Abd El Ghany, Mohammed A.‐M. Salem · 2023
Breast cancer stands as a leading cause of global cancer-related deaths among women, emphasizing the urgency for advancements in early detection and precise diagnosis to enhance patient outcomes. This study addresses the pressing need for improved breast cancer imaging through the integration of deep learning algorithms.This paper specifically investigates the application of U-Net architecture and Federated Averaging (FedAvg) techniques for enhancing ultrasound breast cancer images thereby assisting radiologists in accurate cancer diagnosis. Leveraging two authentic datasets, we conduct experiments varying the number of epochs (50, 100, and 200) for FedAvg with U-Net. Our findings demonstrate promising accuracy values: 89.4716, 89.48, and 89.48, showcasing the effectiveness of our approach.Additionally, we explore the impact of U-Net alone and in conjunction with FedAvg on the segmentation process of ultrasound breast cancer images. This dual-pronged investigation offers insights into the synergistic potential of these methodologies for image enhancement.In conclusion, our study contributes to the ongoing efforts for improved breast cancer imaging. Through the fusion of U-Net and FedAvg, our approach demonstrates significant promise, providing a valuable avenue for advancing medical image analysis in the context of breast cancer diagnosis and treatment planning.