基于定量CT对儿童难治性支原体肺炎的分析研究

Analytical study of refractory mycoplasma pneumonia in children based on quantitative CT

目的:基于计算机肺部感染辅助诊断对难治性儿童支原体肺炎(RMPP)进行CT定量分析,寻找影像学定量分析的敏感指标,并探讨预测RMPP的相关因素。方法:回顾性收集118例肺炎支原体肺炎(MPP)患儿的临床及影像资料,其中难治性支原体肺炎(RMPP)患儿30例,普通支原体肺炎(GMPP)患儿88例。利用计算机肺部感染辅助诊断分析软件对患儿胸部CT进行定量分析。采用χ~2检验或Mann-Whitney U检验比较两组的一般资料及CT定量参数的差异,将组间存在统计学差异的指标行ROC曲线和多因素logistic逐步回归分析。结果:RMPP组较GMPP组年龄较大、住院时间较长、临床症状较重(P<0.05)。CT定量参数中病变区域容积(LV)、病变占全肺体积的百分比(LV%)、病变平均密度(MLD)及病变质量(LQ)均高于GMPP组,低于-300 HU的低密度区百分比(GGO%)低于GMPP组,差异具有统计学意义(P<0.05)。ROC曲线显示LV、LV%、MLD、GGO%及LQ均有助于区分RMPP及GMPP患儿,多因素logistic逐步回归显示LV%>4.2、MLD>-296.36 HU、LQ>21.6 g是RMPP的独立危险因素。结论:计算机肺部感染辅助诊断工具能够识别与定量分析儿童支原体肺炎。LV、LV%、MLD及LQ可用于鉴别RMPP患儿,当LV%>4.2、MLD>-296.36 HU、LQ>21.6 g是儿童RMPP的重要预测指标,为临床治疗提供依据。

Objective:To find the sensitive biomarker of quantitative imaging features and explore the independent risk factors for predicting refractory mycoplasma pneumonia(RMPP) in children,based on quantitative CT analysis by pulmonary infection computerized aid diagnostical tool.Methods:The clinical and imaging data of 118 patients with MPP(including 30 cases of RMPP and 88 cases of GMPP) were collected retrospectively.The chest CTs of children were quantitatively analyzed by computerized aid diagnosis software.The differences in clinical information and quantitative CT parameters between the two groups were compared by using the Chi-square test or Mann-Whitney U test.The indicators with statistical differences between the groups were analyzed by ROC curves and multivariate logistic stepwise regression analysis.Results:The RMPP group was older,had a longer hospital stay and more severe clinical symptoms than the GMPP group(P<0.05).Quantitative CT parameters of lesion volume(LV),lesion percentage of whole lung volume(LV%),lesion mean density(MLD),and lesion quality(LQ) were higher than those in the GMPP group,and the percentage of hypodensity areas below-300 HU(GGO%) was lower than that in GMPP group,with statistically significant difference(P<0.05).ROC curves showed that LV,LV%,MLD,GGO%,and LQ are helpful to distinguish children with RMPP from those with GMPP.Multivariate logistic stepwise regression showed that LV%>4.2,MLD>-296.36 HU,and LQ>21.6 g were independent risk factors for predicting RMPP.Conclusion:The lung infection computerized aid diagnosis tool can accurately identify and quantifies mycoplasma pneumonia in children.LV,LV%,MLD,GGO%,and LQ can be used to identify children with RMPP.LV%>4.2,MLD>-296.36 HU,LQ>21.6 g are important predictors of RMPP in children and provide a basis for clinical treatment.