CT图像纹理分析在新型冠状病毒肺炎临床分型诊断中的应用价值

The application value of CT image texture analysis in the diagnosis of clinical types of COVID-19

目的探讨CT图像纹理分析在新型冠状病毒肺炎临床分型诊断中的应用价值。方法回顾性分析2020年1月22日~3月5日潍坊地区经临床专家确诊的32例新型冠状病毒肺炎患者的影像及临床资料,并依据临床分型标准把32例患者分为普通组(轻型及普通型患者20例)和重症组(重型及危重型患者12例),运用Image J软件分别在各自病灶的最大层面沿病灶边缘手动勾画感兴趣区(ROI),测量其直方图参数及灰度共生矩阵参数,比较两组纹理参数差异,对有统计学意义的指标绘制受试者工作特征(ROC)曲线,计算ROC曲线下面积(AUC),预测纹理分析的效能。结果纹理参数标准差、偏度、峰度、角二阶矩、对比和熵共6组参数差异有统计学意义(P<0.05),其中标准差、角二阶矩和熵3组参数诊断效能较高,AUC分别为0.761,0.860,0.866,采用二元逻辑回归分析将标准差、角二阶矩和熵3组纹理参数联合分析,其中,标准差联合熵及标准差联合角二阶矩AUC分别为0.910,0.904,诊断效能更好。结论CT图像纹理分析在新型冠状病毒肺炎临床分型诊断中具有一定的价值。

Objective To explore the application value of CT image texture analysis in the diagnosis of clinical types of COVID-19.Methods Retrospective analysis of 32 cases of COVID-19 confirmed by clinical experts in Weifang district from 22 January to 5 March 2020,and according to clinical types,32 patients were divided into general groups(20 cases of light and general patients)and severe groups(12 cases of heavy and critical patients),using Image J software manually sketches the interest area(ROI)along the edge of the lesions at the maximum level of their respective lesions,measures their histogram parameters and gray symbous matrix parameters,compares the differences between the two sets of texture parameters,draws the receive operating characteristic(ROC)curve for statistically significant indicators,and calculates the area with curve,which predicts the effectiveness of texture analysis.Results The difference between the standard deviation,bias,peak,angle second-order moment,contrast and entropy of the texture parameters was statistically significant(P<0.05),of which the standard deviation,angle second-order moment and entropy parameters had higher diagnostic efficiency,and the AUC was 0.761.0.860,0.866,using binary logic regression analysis will be standard deviation,angle second-order moment and entropy 3 sets of texture parameters joint analysis,in which the standard deviation joint entropy and standard deviation joint angle second-order moment AUC was 0.910,0.904,better diagnostic performance.Conclusion CT image texture analysis has some value in the diagnosis of clinical types of COVID-19.