常规CT征象及MaZda纹理分析鉴别肺纯磨玻璃结节侵袭性的诊断价值

Value of conventional CT signs and MaZda texture analysis technology in identifying the aggressiveness of pGGO

目的探讨常规CT征象及MaZda纹理分析技术在鉴别肺纯磨玻璃结节(pure ground glass opacity,pGGO)侵袭性的诊断价值。方法选取93例(97个pGGO)肺腺癌患者,分为侵袭前组及侵袭性组。用Mazda软件对CT平扫肺窗图像进行纹理分析:分析方法(直方图、绝对梯度、游程矩阵、共生矩阵、自回归模型和小波转换);提取特征参数方法(费希尔算法、分类误判率+平均相关系数法、相关信息测度);B11判别方法(主成分分析法、线性判别分析、非线性判别分析),得到鉴别pGGO侵袭性的误判率并进行ROC检验,CT及临床资料进行SPSS统计分析。结果Fisher+LDA组合误判率最低,为2/97(2.06%)。筛选得到10个最优纹理参数(均值、五个百分位灰度值、三个逆差距、非零化百分比),P值均<0.01。CT征象鉴别pGGO侵袭性在结节大小、形状、毛刺征、胸膜凹陷征、血管走行关系中差异有统计学意义(P<0.01)。结论常规CT征象及MaZda纹理分析用于鉴别肺pGGO侵袭性是可行的。

Objective To investigate the diagnostic value of conventional CT signs and MaZda texture analysis technology in identifying the aggressiveness of lung pure ground glass opacity(pGGO).Methods Retrospective analysis was studied on 93 patients with 97 pGGO with lung adenocarcinoma lesions.A total of 93 cases were divided into 2 groups:Preinvasion group and Invasive group.Both of the 2 groups were analyzed through Texture analysis software(MaZda)based on CT in order to identify the texture features of lung adenocarcinoma:using different analysis model of histogram,absolute gradient,run-length matrix,co-occurrence matrix,autoregressive model and wavelet transformation;In addition,four methods containing Fisher coefficient,classification error probability with average classification coefficients(POE+ACC)and mutual information(MI)were performed to filter optimal texture features.Different statistic methods,including the principal component analysis(PCA),linear discriminant analysis(LDA),nonlinear discriminant analysis(NDA)and statistical methods such as ROC test,SPSS,were integrated to calculate the minimum misdiagnosis rates(R)of texture parameters using B11 software.Results The lowest R of Fisher+LDA was 2.06%(2/97).We screened out 10 best texture features(mean,five percentile grayscale,three inverse difference moment,percentage nonzero gradient)with lower misdiagnosis(P<0.01)and identifid the Statistical difference between aggressiveness of pGGO and nodule size,shape,speculation,pleural depression sign,and blood vessel running(P<0.01).Conclusion Our finding demonstrated the feasibility of conventional CT signs and MaZda texture analysis in identifying the invasiveness of lung pure ground glass opacity.