Attention Makes Breast Cancer Segmentation in Ultrasound Image Better

According to the International Agency for Research on Cancer (IARC) statistics, starting from the year 2020, breast cancer has officially surpassed lung cancer and becomes the world's leading cancer. Early detection of tumors plays a crucial role in clinical treatment, and ultrasound scanning is a cheap and efficient imaging technique that is widely used. Today, with the development of deep learning, an increasing number of network architectures have demonstrated promising performance in medical image processing, among which UNet stands out as an excellent example. In this paper, we compare the effectiveness of UNet,and UNet with attention mechanism,such as TransUnet and SwinUnet in detecting and segmenting breast cancer lesion area based on ultrasound images. From the experiment, UNet with attention show an excellent performance, especially for the TransUnet, which has the best indicators among all the models.