An Improved Cell nuclei Segmentation Method Based on U-Net++

Segmentation of nuclei plays a crucial role in the field of medical image processing, particularly in the identification and examination of diseasesAlthough the U-Net ++ has shown good results in this task, there are still some limitations, and further improvement is needed to improve its segmentation accuracy. In this study, we investigate the limitations of U-Net++ and propose an enhanced model that incorporates the CBAM attention mechanism and BasicBlock residual block into the original U-Net++ architecture.The CBAM attention mechanism enables adaptive adjustment of both channel and spatial dimensions of the feature map, while the BasicBlock residual block effectively addresses the issues of gradient vanishing and challenging model training in deep networks.To validate the efficacy of the enhanced U-Net++ model, we experimented on a publicly available DSB2018 nucleus dataset. The experimental results demonstrate that the enhanced U-Net++ model achieves a $0.75 \%$ increase in IOU and a $2.13 \%$ improvement in the DICE similarity coefficient, indicating superior performance compared to the original model.