现代电子技术2025，Vol.48Issue(13)：29-35,7.DOI:10.16652/j.issn.1004-373x.2025.13.004
基于改进U-Net的细胞核图像分割网络
A nucleus image segmentation network based on improved U-Net
摘要
Abstract
Deep learning model based on convolutional neural network(CNN)have achieved significant breakthrough in biomedical image segmentation and have been widely applied in practical scenarios.The accuracy of nucleus image segmentation plays a crucial role in pathological diagnosis.However,the existing nucleus segmentation algorithms still suffer from issues such as fuzzy and adherent boundaries,so an image segmentation algorithm based on improved U-Net is proposed.In the model,a triple attention module is utilized to enhance feature focus,and feature fusion module,AG gate module,and lightweight Inception module are incorporated to improve segmentation accuracy.The proposed algorithm was validated on the 2018 Data Science Bowl(DSB2018)dataset.The evaluation metrics including IoU(intersection over union)and DSC reach 81.85%and 90.00%,respectively.Experimental results demonstrate that in comparison with the other segmentation models,the proposed algorithm exhibits significant advantages in terms of the conformity between segmented results and ground truth labels,achieving superior performance.关键词
卷积神经网络/深度学习/细胞核分割/U-Net网络/注意力机制/图像分割Key words
CNN/deep learning/nucleus segmentation/U-Net network/attention mechanism/image segmentation分类
电子信息工程引用本文复制引用
宋文博,祝开艳,刘通,宋维波..基于改进U-Net的细胞核图像分割网络[J].现代电子技术,2025,48(13):29-35,7.基金项目
辽宁省教育厅项目(LJKMZ20221110) （LJKMZ20221110）
