Attention U-Net: A Deep Learning Approach for Breast Cancer Segmentation
Krishna Mridha, Tasnim Sarker, Suborno Deb Bappon, Shahriar Mahmud Sabuj · 2023
One of the main causes of death for women globally is breast cancer. Early discovery can reduce the frequency of early deaths. Breast ultrasound images can help in diagnosis and treatment planning by helping to categories different types of breast cancer. To segment breast cancer pictures in this study, we employed an Attention U-Net. A deep learning model for picture segmentation is called the Attention U-Net. We used a dataset of 780 breast ultrasound pictures divided into three classes-normal, benign, and malignant-to train the Attention U-Net. The mean IoU, precision, recall, and F1 scores for the Attention U-Net were 0.840,0.815,0.848, and 0.831 respectively. These outcomes are comparable to those of other cutting-edge techniques for segmenting breast cancer. The findings of this study indicate that the Attention U-Net is a potential technique for segmenting breast cancer. The great accuracy that the Attention U-Net can attain can assist to increase the precision of diagnosis and treatment planning