基于改进U-Net网络的新冠病毒CT图像分割  

Segmentation of COVID-19 CT Images Based on U-Net Networks

作　　者：陈成 张正 肖迪 CHEN Cheng;ZHANG Zheng;XIAO Di(College of Electrical Engineering and Control Science,Nanjing Tech University,Nanjing Jiangsu 211816,China)

机构地区：[1]南京工业大学电气工程与控制科学学院,江苏南京211816

出　　处：《计算机仿真》2023年第7期154-158,452,共6页Computer Simulation

基　　金：智能机器人与系统高精尖创新中心开放基金([2020]001);江苏省研究生科研与实践创新计划基金项目(SJCX20_0350,KYCX21_1137)。

摘　　要：新冠病毒2019(COVID-19)在过去一年内严重威胁人类生命和健康。全球经济、教育、交通等方面都受到了影响。为了尽快解决新冠病毒引起的问题,快速准确判断人们是否感染是非常重要的。以U-Net为基本框架,采用多尺度特征提取模块,并在反卷积过程中加入位置信息,将图像的全局信息和局部信息充分结合,提出改进模型并应用于COVID-19 CT图像集。上述模型通过深度学习自动分割COVID-19 CT图像的左右肺、病灶和背景四部分。最后对分割结果进行了评价,达到了预期的效果。有助于医务人员快速识别感染部位。Corona Virus Disease 2019(COVID-19)has seriously threatened human life and health in the past year.The global economy,education,transportation,and other aspects have been affected.In order to solve the problems caused by COVID-19 as soon as possible,it is very important to quickly and accurately determine whether people are infected.In this paper,we take U-Net as the basic model,adopt a multi-scale feature extraction module and add position information in the deconvolution process to fully combine the global information and local information of the image.We propose a model suitable for the COVID-19 CT image sets.The model can automatically segment four parts(left and right lung,disease,and background)of COVID-19 CT images through deep learning.The segmentation results are evaluated and the expected results are achieved.It is helpful for medical staff to quickly identify the infection area.

关 键 词：新冠病毒 医学图像分割 多尺度特征 位置信息 

分 类 号：TP183[自动化与计算机技术—控制理论与控制工程]