计算机应用

• 人工智能与仿真 •    下一篇

基于注意力机制的多尺度残差UNet实现乳腺癌灶分割

罗圣钦1,陈金怡1,李洪均2   

  1. 1. 南通大学信息科学技术学院
    2. 南通大学
  • 收稿日期:2021-06-04 修回日期:2021-06-23 发布日期:2021-11-09 出版日期:2021-11-09
  • 通讯作者: 罗圣钦

Multiscale residual UNet based on attention mechanism to realize breast cancer lesion segmentation

  • Received:2021-06-04 Revised:2021-06-23 Online:2021-11-09 Published:2021-11-09

摘要:

摘 要: 针对乳腺癌灶在磁共振成像(MRI)中呈现大小不一、边界模糊等特点,为避免误割并提高分割精度,提出一种基于注意力机制的多尺度残差U网络分割算法。首先,利用多尺度残差单元替换UNet在下采样过程中的相邻两个卷积块以加强对形态大小差异的关注。接着,在上采样阶段使用跨层的注意力引导网络对重点区域的关注,避免造成对健康组织的误分割。最后,引入空洞空间金字塔池化作为分割网络的桥接模块以强化对病灶的表征能力。与UNet相比,该算法在Dice系数、交并比(IOU)、特异度(SP)、准确度(ACC)等指标上分别提升了2.72%、2.93%、5.15%、0.05%。实验结果表明,所提算法能够提高癌灶分割精度,有效降低影像诊断的假阳性率。

Abstract:

Abstract: Concern the characteristics of breast cancer in magnetic resonance imaging(MRI), such as different shapes and sizes, fuzzy boundaries and so on, an algorithm based on multiscale Res-U network with attention mechanism was proposed in order to avoid missegmentation and improve segmentation accuracy. Firstly, the multiscale residuals were used to replace two adjacent convolution blocks in the subsampling process of UNet, so that the network could pay more attention to the difference of morphology and size. Then, in the up-sampling stage, layer-crossed attention was used to guide the network to focus on the key regions, avoiding the missegmentation of healthy tissues. Finally, in order to enhance the ability of representing the lesions, the atrous spatial pyramid pooling was introduced as a bridging module to the network. Compared with UNet, the proposed algorithm improved the Dice coefficient, Intersection over Union (IOU), Specificity (SP) and Accuracy (ACC) by 2.72%, 2.93%, 5.15% and 0.05%, respectively. The experimental results show that the algorithm can improve the segmentation accuracy of lesion and effectively reduce the false positive rate of imaging diagnosis.

中图分类号: