当前位置>主页 > 期刊在线 > 计算机技术 >

计算机技术21年13期

基于改进 U-Net 的全心脏 CT 图像分割
陈秋叶,韦瑞华,石璐莹,吴甜,刘海华
(中南民族大学,湖北 武汉 430074)

摘  要:针对 CT 图像中全心脏结构复杂度高、分割不完整及分割精度低等问题,文章提出了一种改进 U-Net 的全心脏分割方法。根据全心脏结构形态特点,文章将多并行尺度特征融合模块引入 U-Net 网络的编码层,并在 U-Net 网络的跳层连接中加入了注意力机制。文章利用 MM-WHS 数据集将改进的全心脏分割算法在中南民族大学认知科学实验室中进行了一系列的全心脏分割实验。实验结果显示,文章提出的算法分割相似度达到 88.73%,提高了全心脏结构的分割准确率。


关键词:全心脏 CT 图像分割;改进 U-Net 网络;多并行尺度特征融合;注意力机制



DOI:10.19850/j.cnki.2096-4706.2021.13.019


基金项目:国家自然科学基金项目资助项 目(61773409)


中图分类号:TP391                                      文献标识码:A                                         文章编号:2096-4706(2021)13-0076-05


Whole-heart CT Image Segmentation Based on Improved U-Net

CHEN Qiuye, WEI Ruihua, SHI Luying, WU Tian, LIU Haihua

(South-Central Minzu University, Wuhan 430074, China)

Abstract: Aiming at the problems of high structural complexity, incomplete segmentation and low segmentation accuracy of whole heart in CT images, the paper proposes a whole-heart segmentation method based on improved U-Net. According to the structural and morphological characteristics of the whole heart, multi-parallel scale feature fusion module is introduced into the coding layer of U-Net network, and attention mechanism is added into the skipping layer of U-Net network. A series of whole heart segmentation experiments are carried out in the Cognitive Science Laboratory of South-Central University for Nationalities using the improved whole heart segmentation algorithm based on the MM-WHS dataset. Experimental results show that the similarity of the proposed algorithm reaches 88.73%, which improves the segmentation accuracy of the whole heart structure.

Keywords: whole heart CT image segmentation; improved U-Net network; multi parallel scale feature fusion; attention mechanism


参考文献:

[1] 胡盛寿,高润霖,刘力生,等 .《中国心血管病报告 2018》概要 [J]. 中国循环杂志,2019,34(3):209-220.

[2] ZHUANG X. Challenges and methodologies of fully automatic whole heart segmentation:a review [J].Journal of Healthcare Engineering,2013,4(3):371-408.

[3] ZHUANG X,SONG J,ZHAN S,et al. A registration and atlas propagation based framework for automatic whole heart segmentation of CT volumes [C]//SPIE Medical Imaging. Orlando: SPIE,2013.

[4] PAYER C,ŠTERN D,BISCHOF H,et al. Multi-label whole heart segmentation using CNNs and anatomical label configurations [C]// STACOM 2017:Statistical Atlases and Computational Models of the Heart. ACDC and MMWHS Challenges. Quebec:Springer,Cham, 2018:190-198.

[5] YANG X,BIAN C,YU L,et al. Hybrid loss guided convolutional networks for whole heartparsing [C]//STACOM 2017: Statistical Atlases and Computational Models of the Heart. ACDC and MMWHS Challenges. Quebec:Springer,Cham,2017:215-223.

[6] XU Z,WU Z,FENG J. CFUN:Combining Faster R-CNN and U-net Network for Efficient Whole Heart Segmentation [J/OL]. arXiv.org,2018,1(2018-12-12).https://arxiv.org/abs/1812.04914v1.

[7] YU F,KOLTUN V. Multi-Scale Context Aggregation by Dilated Convolutions [J/OL].arXiv.org,2016,3[2021-06-01].https:// arxiv.org/abs/1511.07122.

[8] 叶承钦 . 基于编解码结构的全心脏 CT 图像分割 [D]. 哈 尔滨:哈尔滨工业大学,2019.

[9] ZHANG S,FU H,YAN Y,et al. Attention Guided Network for Retinal Image Segmentation [C]//MICCAI 2019:Medical Image Computing and Computer Assisted Intervention – MICCAI 2019. Shenzhen:Springer,Cham,2019:797-805.

[10] HABIJAN M,GALI I,LEVENTI H,et al. Whole heart segmentation using 3D FM-Pre-ResNet encoder–decoder based architecture with variational autoencoder regularization [J].Applied Sciences,2021,11(9):3912.

[11] LIU T,TIAN Y,ZHAO S,et al. Automatic Whole Heart Segmentation Using a Two-Stage U-Net Framework and an Adaptive Threshold Window [J].IEEE Access,2019,7:83628-83636.

[12] 付殿臣 . 跨模态心脏医学图像域自适应分割算法研究 [D]. 长春:长春工业大学,2021.

[13] HABIJAN M,LEVENTI H,GALI I,et al. Neural network based whole heart segmentation from 3D CT images [J].International journal of electrical and computer engineering systems,2020,11(1):[2021-05-31].http://www.etfos.unios.hr/ ijeces/papers/neural-network-based-whole-heart-segmentation-from3d-ct-images/.

[14] RONNEBERGER O,FISCHER P,BROX T. U-Net: Convolutional Networks for Biomedical Image Segmentation [C]//MICCAI 2015:Medical Image Computing and Computer-Assisted Intervention – MICCAI 2015.Munich:Springer,Cham,2015:234-241.


作者简介:陈秋叶(1995—),女,汉族,河南开封人,硕士 研究生在读,研究方向:视觉认知与医学图像处理;通讯作者:刘 海华(1966—),男,汉族,湖北孝感人,教授,博士,研究方向: 视觉认知计算与医学图像处理。