Optimization of breast medical image segmentation algorithm based on U-Net
A study of the traditional U-Net network for breast cancer segmentation reported that the network is effective in im-age segmentation,however,the complex ultrasound patterns and the adhesion of the surrounding tissue cells make it difficult for these simple frameworks to achieve the desired segmentation results on breast ultrasound images.Therefore the U-Net structure was improved by introducing the polarized self-attention mechanism(PSA)as well as self-calibrated convolution(SCConv)to de-sign a better network architecture.PSA dynamically adjusts the weights of the feature map by at the pixel level and SCConv is able to dynamically adjust the weights of the convolution kernel to better capture and represent the local and global information of the in-put image.
万方数据
