Madr-net: Multi-level Attention Dilated Residual Neural Network For Segmentation Of Medical Images

This article discusses a new deep learning framework, called MADR-Net, designed to improve the performance of medical image segmentation. It uses a combination of U-Net encoder/decoder backbone, multi-level residual blocks, and channel-spatial attention blocks to capture both global and local features. The proposed algorithm has been extensively validated on four challenging medical image segmentation tasks and has shown significant improvement compared to other state-of-the-art architectures.