基于AI定量检测新型冠状病毒肺炎胸部CT演变特征分析  

Quantitative Detection of Chest CT Evolution Characteristics of COVID-19 Based on Artificial Intelligence(AI)

作　　者：黄晓旗 王莉[1] 史柯 梁玉栋 刘亚良 张静平[5] 金晨望[5] 郭佑民[1,5] HUANG Xiao-qi;WANG li;SHI Ke;LIANG Yu-dong;LIU Ya-liang;ZHANG Jin-ping;JIN Chen-wang;GUO You-min(Department of Radiology,The Affiliated Hospital of Yan’an University,Yan’an 716000,Shaanxi Province,China;Department of Radiology,The First Affiliated Hospital of Xi’an Jiaotong University,Xi’an 710061,Shaanxi Province,China;Department of Radiology,Ankang people's Hospital,Ankang 725000,Shaanxi Province,China;Department of CT&MR imaging diagnostics,Weinan Central Hospital,Weinan 714000Shaanxi Province,China;Department of Radiology,Hanzhong Central Hospital,Hanzhong 723000,Shaanxi Province,China)

机构地区：[1]延安大学附属医院影像科,陕西延安716000 [2]安康市人民医院放射科,陕西西安710061 [3]渭南市中心医院CT/MR影像诊断科,陕西安康725000 [4]汉中市中心医院放射科,陕西渭南714000 [5]西安交通大学第一附属医院放射科,陕西汉中723000

出　　处：《中国CT和MRI杂志》2022年第11期55-57,共3页Chinese Journal of CT and MRI

基　　金：陕西省教育厅2020年度突发公共卫生安全专项科学研究计划(20JG040,20JG039,20JG019)。

摘　　要：目的探讨人工智能(AI)定量检测在新型冠状病毒肺炎(COVID-19)不同分型中胸部CT动态变化特征的价值。方法回顾性分析2020年1月至2020年5月期间多家医疗中心确诊的95例COVID-19非重症患者和22例COVID-19重症患者的连续动态CT扫描结果及首次临床症状、实验室检查结果。使用AI技术自动分割COVID-19病灶得到CT定量结果。采用独立样本t检验或非参数检验比较两组间定量指标的差异。对发病后天数与CT定量数据进行三次多项式曲线回归函数拟合,得到变化曲线。结果重症组患者除GGO%小于非重症组外,剩余CT定量指标均大于非重症组(P﹤0.05)。非重症组CT定量指标约第10~11天达到峰值,10~30天内迅速好转,30~40天基本消散。重症组发病高峰迟于非重症组1~2天,病变吸收速度远低于非重症组,并且消失时间延长。结论AI定量检测有助于准确识别COVID-19肺炎胸部CT的动态演变过程。不同定量指标能够客观地评估肺内病变性质,为疾病的转归提供客观依据。Objective To investigate the value of quantitative by artificial intelligence(AI)detection in the dynamic characteristics of chest CT in different types of coronavirus disease 2019(COVID-19).Methods The continuous dynamic CT scan results,the first clinical symptoms and laboratory examination results of 95 patients with non-severe COVID-19 and 22 patients with severe COVID-19 diagnosed in many medical centers from January 2020 to May 2020 were analyzed retrospectively.The quantitative results of CT were obtained by automatically segmenting COVID-19 lesions by AI technique.The differences of quantitative indexes between the two groups were compared by independent sample t-test or nonparametric test.The days after onset and CT quantitative data were fitted by cubic polynomial curve regression function to estimate the pattren of CT quantitative data with the time after onset.Results Except the GGO%in severe group was lower than that in non-severe group,the other CT quantitative indexes were higher(P﹤0.05).In the non-severe group,the quantitative index of CT reached the peak at about 10-11 days,improved rapidly in 10-30 days,and basically dissipated in 30-40 days.The peak of onset in the severe group was 1-2 days later than that in the non-severe group,the absorption rate of lesions was much lower than that in the non-severe group,and the disappearance time was prolonged,the absorption rate of lesions was much lower than that of the non-severe group,and the disappearance time was prolonged.Conclusion Quantitative AI is helpful to accurately identify the dynamic evolution of CT in COVID-19.Different quantitative indexes can objectively evaluate the nature of pulmonary lesions and provide objective basis for the prognosis of the disease.

关 键 词：新型冠状病毒肺炎(COVID-19) 人工智能 体层摄影术 X线计算机 

分 类 号：R563.1[医药卫生—呼吸系统] R816.4[医药卫生—内科学]