For the past several years, semantic segmentation method based on deep learning, especially Unet, have achieved tremendous success in medical image processing. The U-shaped topology of Unet can well solve image segmentation tasks. However, due to the limitation of traditional convolution operations, Unet cannot realize global semantic information interaction. To address this problem, this paper proposes RT-Unet, which combines the advantages of Transformer and Residual network for accurate medical segmentation. In RT-Unet, the Residual block is taken as the image feature extraction layer to alleviate the problem of gradient degradation and obtain more effective features. Meanwhile, Skip-Transformer is proposed, which takes Multi-head Self-Attention as the main algorithm framework, instead of the original Skip-Connection layer in Unet to avoid the influence of shallow features on the network's performance. Besides, we add attention module at the decoder to reduce semantic differences. According to the experiments on MoNuSeg data set and ISBI_2018cell data set, RT-Unet achieves better segmentation performance than other deep learning-based algorithms. In addition, a series of further ablation experiments were conducted on Residual network and Skip-Transformer, which verified the effectiveness and efficiency of the proposed methods in this paper.

Currently, deep learning has become more and more mature in the field of medical image segmentation. Through using the computer, the deep learning models established can completely help doctors to perform medical image segmentation. Most of the current deep learning models are based on Unet. The U-shaped structure and the skip connection layer of Unet can effectively achieve precise image segmentation. However, for complicated images, the network structure of Unet is not sufficient enough. In response to this problem, some scholars have designed Unet++ by adding a denser skip connection layer to U-Net. Compared to U-Net, Unet++ is more effective in dealing with complex images, but it has drawbacks in many aspects, and there is still a large loss of eigenvalues in the skip connection and up-sampling processes. To address these issues, this paper uses the channel and attention mechanism to improve the Unet++ model to obtain better image segmentation efficiency and accuracy. Meanwhile, based on Unet++, this paper designs a new model called CA-Unet++. The proposed model uses the channel module and the attention module to solve the eigenvalues loses in the long-distance skip connection process and the up-sampling process, respectively. The experimental results and data analysis shows that our proposed CA-Unet++ can achieve better performance in medical computed tomography scan image segmentation.