摘要
目的:利用人工智能(artificial intelligence,AI)技术对COVID-19患者的胸部影像学资料进行数字化分析,并总结其特点及分析AI技术的优势。方法:应用依图肺部智能AI辅助系统对十堰市人民医院核酸检测阳性且经临床多学科会诊确诊的66例COVID-19患者的CT影像学资料及临床资料进行回顾性分析。利用GE Light⁃Speed16MSCT进行胸部扫描,将影像学资料导入AI系统进行量化。总结新冠肺炎患者的CT征象,包括磨玻璃影、实变影等,并对AI系统构筑的密度直方图等进行分析。采用SPSS 23.0软件进行数据分析,对左右肺的磨玻璃、实变及总病灶体积进行对比分析。结果:COVID-19患者主要的影像学表现为双肺散在分布磨玻璃密度影伴部分肺组织实变;AI辅助诊断系统较人工阅片具有优势,可以定量病灶体积并能更好地识别病灶累计的肺叶。COVID-19患者右肺病灶体积、磨玻璃影病灶体积及实变病灶体积均大于左肺,差异有统计学意义(P<0.05);并且AI辅助系统构建的密度直方图可统计区间内各种类型病灶的像素数量,对监测病情发展起到了一定作用。结论:AI技术的检测效率、能力及准确率均优于人工识别,AI技术可以通过量化病灶以及构建直方图等形式,更加直观地总结病灶的分布特点并分析病情的进展情况,因此AI技术对COVID-19的诊断具有重要的临床价值。
Objective The chest imaging data of COVID-19 patients were digitally analyzed using artificial intelligence(AI)technology,and the characteristics and advantages of AI technology were summarized.Methods The CT imaging data and clinical data of 66 patients with COVID-19 who were positive for nucleic acid test and confirmed by clinical multidiscipli⁃nary consultation in Shiyan Renmin Hospital were retrospectively analyzed with Yitu Lung Intelligent AI-assisted system.The chest scan was performed with GE LightSpeed16MSCT,and the imaging data were imported into the AI system for quantification.CT signs of COVID-19 patients were summarized,including ground glass shadow,consolidation shadow,etc.The density histogram constructed by AI system was observed and summarized.SPSS 23.0 software was used for data a⁃nalysis,and the ground glass,consolidation and total lesion volume of left and right lungs were compared and analyzed.Results The main imaging findings of COVID-19 patients were scattered ground-glass density shadow with partial lung consolidation in both lungs.The AI-assisted diagnosis system has advantages over manual image reading,which can quanti⁃fy the volume of lesions and better identify the accumulated lobes of lesions.The lesion volume of right lung in COVID-19 patients was larger than that of left lung,and the difference was statistically significant(P<0.05).The volume of right lung lesion,ground glass shadow lesion and consolidation lesion in patients with cowid-19 were larger than those in left lung,and the difference was statistically significant(P<0.05).In addition,the density histogram constructed by the AI-assisted system can count the number of pixels of various types of lesions within the interval,which plays a certain role in monitoring the development of the disease.Conclusion The detection efficiency,ability and accuracy of AI technology are better than that of manual recognition.AI technology can be more direct by quantifying lesions and constructing histogram.Therefore,AI technology has im
作者
刘书铭
曾照军
敖峰
谢兴武
贺欢
陈光斌
LIU Shu-ming;ZENG Zhao-jun;AO Feng;XIE Xing-wu;HE Huan;CHEN Guang-bin(Postgraduate Training Base,Renmin Hospital,Jinzhou Medical University;Department of Radiography,Renmin Hospi-tal,Hubei University of Medicine;Institute of Radiological Imaging,Renmin Hospital,Hubei University of Medicine,Shiy-an,Hubei 442000,China)
出处
《湖北医药学院学报》
CAS
2021年第3期273-276,280,共5页
Journal of Hubei University of Medicine
基金
十堰市科技局新冠肺炎科研立项(2020XGFYZR18)。
