A dual-path U-Net for pulmonary vessel segmentation method based on lightweight 3D attention

In recent years, pulmonary vessel segmentation has aroused widespread interest in medical image analysis. However, the structure of pulmonary vessels is complex and CT images have a lot of noise. Therefore, it is a difficult task to extract pulmonary vessels accurately. In terms of pulmonary vessel segmentation, medical image segmentation methods based on deep learning only utilize single volume data in CT image, but cannot fully fuse multiple volume data, resulting in the accuracy of pulmonary vessel segmentation is low. In order to fully utilize the complementary advantages of multiple volume data, we propose a DS-ResUNet to segment pulmonary vessels in multi-view. The DS-ResUNet uses feature fusion module to fuse 2  mm and 5 mm volume data and makes full use of the detailed vessel textures in 2 mm volume data and coarse vessel structures in 5 mm volume data. By this multi-view fusion module, the accuracy of pulmonary vessel segmentation can be effectively improved. In addition, in order to strengthen the spatial weight of vessels and reduce the model parameters, we design a lightweight 3D axial attention module by separable convolution. To confirm the improved performance, we design some comparison experiments with the state-of-the-art segmentation methods. As a result, our DS-ResUNet has a better performance than other state-of-the-art methods on pulmonary vessel segmentation, but also has fewer parameters.