Cfformer: Cross Cnn-Transformer Channel Attention and Spatial Feature Fusion for Improved Segmentation of Low-Quality Medical Images

Abstract

Medical image segmentation plays an important role in computer-aided diagnosis. Existing methods mainly utilize spatial attention to highlight the region of interest. However, due to the constraints of medical imaging devices, many images are sampled with low quality. They are difficult to provide sufficient texture information, reducing the efficiency of spatial computation. To address this issue, we propose a hybrid CNN-Transformer model that leverages effective channel feature extraction to capture contextual information, called CFFormer. The proposed architecture contains two key components: the Cross Feature Channel Attention (CFCA) module and the X-Spatial Feature Fusion (XFF) module. The model incorporates dual encoders, with the CNN encoder focusing on capturing local features and the Transformer encoder modeling global features. The CFCA module filters and facilitates interactions between the channel features from the two encoders, while the XFF module effectively reduces the significant semantic information differences in spatial features, enabling a smooth and cohesive spatial feature fusion. We evaluate our model across eight datasets covering five modalities to test its generalization capability. Experimental results demonstrate that our model outperforms current state-of-the-art methods, with particularly superior performance on datasets characterized by blurry boundaries and low contrast.

Note:
Funding declaration: This work is partially supported by the NSFC project (UNNC Project ID B0166), and Yongjiang Technology Innovation Project (2022A-097-G).

Conflict of Interests: The author declares that there are no financial interests or personal relationships related to this article that could influence the results and conclusions of the study.

Keywords: Medical image segmentation, image segmentation, deep learning, Hybrid CNN-Transformer Model