论文部分内容阅读
目的 评估计算机辅助定量测量的可重复性,探讨计算机辅助定量分析在不同浸润性肺腺癌中的鉴别诊断价值.方法 回顾性分析本院病理证实的328例肺腺癌患者,不典型腺瘤样增生(AAH)/原位癌(AIS) 56例、微浸润腺癌(MIA) 58例、浸润性腺癌(IAC)214例.采用计算机辅助软件对薄层CT图像进行半自动逐层分割,自动生成体积、面积、最大横截面长径、最大横截面短径、最大CT值、最小CT值、平均CT值及通过计算可得平均直径、质量.采用组内相关系数检验分割结果的一致性,采用Kruskal-wallis H检验对上述参数进行差异性检验,最后进行判别分析和有序Logistic回归分析筛选出独立预测因素,并构建模型.结果 上述9个定量参数除最小CT值外,在不同浸润性的肺腺癌间均有统计学差异(P＜0.05),观察者内组内相关系数(ICC)分别为0.997、0.970、0.927、0.849、0.995、0.821、0.989、0.932、0.999,观察者间ICC为0.935、0.916、0.839、0.758、0.949、0.696、0.914、0.867、0.973.判别分析筛选出面积、最大横截面长径及最大CT值用于构造判别函数;有序Logistic回归筛选出的预测因素为最大横截面长径和平均CT值(P＜0.05).结论 计算机辅助定量分析具有良好的观察者内和观察者间一致性,最大横截面长径和平均CT值可作为鉴别磨玻璃密度型肺腺癌浸润性的影像学独立预测因素.“,”Objective To evaluate the reproducibility of computer-aided quantitative measurement and to explore the value of computer-aided quantitative analysis in diagnosing the invasiveness of lung adenocarcinoma.Methods 328 fG-GNs that were pathologically confirmed as lung adenocarcinoma (56 cases of AAH/AIS,58 cases of MIA and 214 cases of IAC) were analyzed retrospectively.The border of the nodule was semi-automatically depicted slice by slice with computeraided software,and the software automatically generated volume,area,maximum cross-sectional length,maximum cross-sectional short diameter,maximum CT value,minimum CT value,mean CT value.Then,mean diameter and mass of fGGNs were calculated.The agreement of the measurement results was evaluated by using intraclass correlation coefficient.The parameters were compared by Kruskal-Wallis H test.Finally,discriminant analysis and ordinal Logistic regression analysis were used to select independent predictors and construct predictive model.Results The above nine quantitative parameters except for minimum CT value were statistically different among three pathological types of AAH/AIS,MIA and IAC (P＜ 0.05).ICC of intraobserver were 0.997、0.970、0.927、0.849、0.995、0.821、0.989、0.932、0.999,ICC of interobserver were 0.935、0.916、0.839、0.758、0.949、0.696、0.914、0.867、0.973.Discriminant analysis selected area,maximum cross-sectional length and maximum CT value to construct discriminant function.Ordinal Logistic regression showed maximum cross-sectional length and mean CT value were predictors(P ＜ 0.05).Conclusion The intra-observer and inter-observer reproducibility of computer-aided quantitative analysis are excellent.Maximum eros-sectional length and mean CT value can be regarded as independent predictors of the invasiveness of lung adenocarcinoma manifesting as ground glass nodule.