Nucleus image segmentation method based on attention mechanism and edge optimization

With the rapid development of deep learning, image segmentation techniques based on deep learning have been widely applied to biomedical image analysis. However, due to the high variability in the shape, size, and color of cell nuclei, as well as noise and blurring during image acquisition, achieving precise segmentation of nucleus images remains challenging. To address these issues, this article proposes a nucleus image segmentation method based on an attention mechanism and edge optimization. By incorporating an efficient shuffle attention mechanism, the method captures channel information while retaining spatial information. Using the Canny operator to precisely extract edge details, an additional parallel auxiliary branch is introduced to extract low-level semantic information. After feature fusion and further optimization, the segmentation edges' clarity and accuracy are enhanced. Experimental results show that the MaskR-CNN with the addition of shuffle attention mechanism and edge optimization, demonstrates significant advantages in various metrics, effectively enhancing the feature extraction and detail retention capabilities for nucleus segmentation, mitigating under-segmentation or over-segmentation issues, and improving the accuracy of nucleus segmentation.