E-Transunet: Enhanced Transunet for Medical Image Segmentation

With the wide application of Transformer in the field of computer vision, the structure combining Convolutional Neural Network (CNN) and Transformer represented by TransUNet demonstrates competitive ability in medical image segmentation. However, due to the inherent limitations of convolutional operations, the encoder in TransUNet is deficient in capturing multi-scale information. In this paper, we propose the Enhanced TransUNet, where an Enhanced Res2Net module is integrated into its encoder to enhance the feature extraction capability. Specifically, the Interaction Block of Enhanced Res2Net module adjusts the fusion of neighboring feature map subsets in Res2Net using an attention mechanism. Experimental results on the Synapse and ACDC datasets demonstrate the strong competitiveness of our approach in medical image segmentation tasks.