基于AI定量分析新型冠状病毒感染不同变异株肺部影像学特征  

Quantitative analysis of pulmonary imaging features of different variants of SARS-CoV-2 infection based on AI

作　　者：王雄慧 呼甜 阴玮灵 王莉[2] 任涛[1] 潘娟[3] 闫军[1] WANG Xionghui;HU Tian;YIN Weiling;WANG Li;REN Tao;PAN Juan;YAN Jun(Department of Radiology,Affiliated Hospital of Yan′an University,Yan′an 716000,China;Department of Respiratory Medicine,Affiliated Hospital of Yan′an University,Yan′an 716000,China;Department of Geriatrics,Yan′an People′s Hospital,Yan′an 716000,China)

机构地区：[1]延安大学附属医院影像科,陕西延安716000 [2]延安大学附属医院呼吸内科,陕西延安716000 [3]延安市人民医院老年病科,陕西延安716000

出　　处：《延安大学学报（医学科学版）》2023年第2期8-13,共6页Journal of Yan'an University:Medical Science Edition

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)。

摘　　要：目的探讨人工智能(artificial inteligence,AI)定量检测在新型冠状病毒感染(coronavirus disease 2019,COVID-19)不同变异株中胸部CT动态变化特征的价值。方法回顾性收集2020年1月至2023年1月期间多家医疗机构确诊的219例COVID-19患者临床资料和多次胸部CT资料。使用AI软件对COVID-19患者出现症状后不同扫描时间的胸部CT图像进行检测分析,得到肺部感染的定量结果。采用卡方检验、独立样本t检验或MannWhitney U检验比较两组间指标的差异。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,判断CT定量数据随发病后时间的变化规律。结果德尔塔毒株感染组的定量CT指标病灶体积(lesion volume,LeV/mL)、病灶占肺体积的百分比(percentage of lesion,LeV%)、平均病灶密度(mean lesion density,MLeD/HU)、异质性和病灶质量(lesion mass,LM/g)均小于奥密克戎毒株感染组(P<0.05)。奥密克戎毒株感染组LeV、LeV%及LM约在发病第8~10天达到高峰,之后迅速下降;德尔塔毒株感染组发病高峰迟于奥密克戎毒株感染组3~4 d,大约在第40天基本消散。奥密克戎毒株感染组与德尔塔毒株感染组在发病初期MLeD达顶峰,但德尔塔毒株感染组病变吸收速度远低于奥密克戎毒株感染组。奥密克戎毒株感染组异质性稍高于德尔塔毒株感染组,德尔塔毒株感染组异质性在10 d后逐渐下降,下降时间远大于奥密克戎毒株感染组。结论奥密克戎毒株较德尔塔毒株肺部感染影像学表现更重,但病变进展及消散迅速。AI定量检测有助于准确识别COVID-19不同变异株胸部CT的动态演变过程,为观察疾病的演变规律及治疗提供客观依据。Objective To investigate the value of artificial intelligence(AI)quantitative detection in characterizing dynamic changes in chest CT in different variants of coronavirus disease 2019(COVID-19).Methods Clinical data and multiple chest CT of 219 patients with COVID-19 diagnosed between January 2020 and January 2023 at multiple medical institutions were retrospectively collected.The chest CT images of COVID-19 patients with different scanning time after the onset of symptoms were detected and analyzed by AI software,and the quantiative index result of pulmonary infection were abtained,independent samples t-test or Mann-Whitney U test were used to compare the differences in indicators between the two groups.A cubic polynomial curve regression function was fitted to the days after onset and the quantitative CT data to determine the pattern of changes in quantitative CT data with time after onset.Results The quantitative CT indices of lesion volume(LeV/mL),percentage of lesion to lung volume(LeV%),mean lesion density(MLeD/HU),heterogeneity and lesion mass in the Delta strain-infected group,(LM/g)were smaller than those of the Omicron strain-infected group(P<0.05).The peak of LeV,LeV%,and LM in the Omicron group peaked at about day 8~10 of disease onset and then declined rapidly;the peak of disease in the Delta group was 3~4 days later than in the Omicron group and largely dissipated by about day 40.The MLeD peaked at the beginning of the disease in the Omicron-infected and Delta-infected groups,but the rate of lesion uptake was much lower in the Delta-infected group than in the Omicron-infected group.Heterogeneity was slightly higher in the Omicron-infected group than in the Delta-infected group,and heterogeneity gradually decreased after 10 days in the Delta-infected group,with a much longer decrease than in the Omicron-infected group.Conclusion The Omicron strain has more severe imaging of the lung infection than the Delta strain,but the lesions progress and dissipate rapidly.The AI quantitative test helps to accura

关 键 词：新型冠状病毒感染 肺炎 定量CT 曲线拟合 

分 类 号：R445.3[医药卫生—影像医学与核医学]