福建电脑2025，Vol.41Issue(11)：42-45,4.DOI:10.16707/j.cnki.fjpc.2025.11.008
注意力驱动的乳腺超声肿瘤精准分割研究
Attention-Driven Precise Segmentation of Breast Ultrasound Tumors
李满堂1
作者信息
- 1. 上海第二工业大学计算机与信息工程学院 上海 201209
- 折叠
摘要
Abstract
This study proposes a tumor segmentation method based on improved Unet++to address the issue of insufficient segmentation accuracy caused by blurry tumor edges and variable shapes in breast ultrasound images.By embedding SE attention modules in skip connections,the correlation learning between feature channels is strengthened,and a mixed loss function combining cross entropy and Dice coefficient is designed to optimize the model training process.Experiments on the publicly available BUSI dataset show that the Dice coefficient of this method reaches 0.891,which is 5.9%higher than the benchmark model,and particularly demonstrates significant advantages in the segmentation of tumor microstructures.The research results verify the effectiveness and progressiveness of the proposed method in breast ultrasound image segmentation.关键词
乳腺肿瘤分割/注意力机制/深度学习/混合损失函数Key words
Breast Tumor Segmentation/Attention Mechanism/Deep Learning/Mixed Loss Function分类
信息技术与安全科学引用本文复制引用
李满堂..注意力驱动的乳腺超声肿瘤精准分割研究[J].福建电脑,2025,41(11):42-45,4.
