【摘要】 针对传统基于CNN和基于Transformer的语义分割方法各自的局限性，提出了一种基于CNN和Transformer的双分支混合模型。该模型通过多注意力Transformer模块捕捉单个样本的上下文信息及不同医学图像之间的关系，有效地兼顾局部和全局特征的建模。此外，设计了双分支交叉融合模块，有效融合来自两个不同分支的特征，提高了模型性能。实验结果表明，模型在Synapse和ACDC数据集上的DSC指标分别为81.31%和91.44%，相较于现有的先进算法，模型在医学图像分割任务中展现了更强的性能，验证了该方法在医学图像处理中的有效性。更多还原

【Abstract】 To address the limitations of traditional CNN-based and Transformer-based semantic segmentation methods, a dualbranch hybrid model based on CNN and Transformer was proposed. A multi-attention Transformer module was utilized in this model to capture contextual information of individual samples and relationships between different medical images, enabling an effective balance between the modeling of local and global features. Additionally, a dual-branch cross-fusion module was designed to effectively combine features from the two different branches, enhancing the model’s performance. Experimental results show that the proposed model achieves DSC scores of 81. 31% and 91. 44% on the Synapse and ACDC datasets, respectively. Compared to existing advanced algorithms, the proposed model demonstrates superior performance in medical image segmentation tasks, validating its effectiveness in medical image processing.更多还原