COVID-19胸部CT结构化报告临床应用价值评价

Evaluation of the clinical application value of structured chest CT report of COVID-19 pneumonia

目的:探究新型冠状病毒肺炎(COVID-19)胸部CT结构化报告临床诊断应用价值。方法:回顾性分析310例疑似COVID-19患者胸部CT及临床资料。根据纳入标准最终共纳入253例患者,阳性组203名,阴性组50名。对胸部CT进行诊断分型,并记录CT征象。以模式1(CT诊断分型)、模式2(CT诊断分型+CT征象)、模式3(CT诊断分型+基本信息+流行病学史+临床及实验室检查)三种模式进行二元逻辑回归分析,采用ROC曲线及曲线下面积(AUC)评价三种模型诊断准确性。结果:逻辑回归分析显示三种模式诊断敏感度分别为95.6%、96.0%和95.2%,特异度分别为34.0%、48.9%和32.7%,符合率分别为83.4%、87.0%、82.2%,AUC为0.768、0.895和0.812。结论:COVID-19胸部CT结构化报告能较为准确地诊断COVID-19肺炎,敏感度较高,但缺乏特异度,结合CT征象可进一步提高诊断准确性,特别是提高诊断特异度。

Objective:To examine the clinical value of chest CT structured report for coronavirus disease 2019(COVID-19)pneumonia.Methods:The chest CT imaging features and clinical data of 310 suspected patients for COVID-19 pneumonia were retrospectively analyzed.Finally,a total of 253 patients were included,203 in the positive group and 50 in the negative group.The CT imaging features were recorded and divided into several categories.Binary logic regression analysis was performed in three modes:mode 1(CT classification),mode 2(CT classification and CT imaging features),mode 3(CT classification,basic information,epidemiological history,clinical and laboratory examination),and the diagnostic accuracy of the three models were evaluated by ROC curve and area under the curve(AUC).Results:Binary logistic regression revealed that the sensitivities of the three models were 95.6%,96.0%,and 95.2%,specificities were 34.0%,48.9%,and 32.7%,accuracies were 83.4%,87.0%,and 82.2%,and the AUC were 0.768,0.895 and 0.812,respectively.Conclusion:Structural chest CT report of COVID-19 pneumonia can accurately diagnose COVID-19 pneumonia,with high sensitivity but lack of specificity.Combined with CT imaging features,the diagnostic accuracy would be further improved,especially for the specificity.