纹理分析T1WI对比增强图像分级脑胶质瘤

目的探讨纹理分析T1WI对比增强图像在分级脑胶质瘤中的价值。 方法回顾性分析2014年10月至2019年10月在扬州大学附属医院，经手术病理证实的29例低级别胶质瘤(LGG)和63例高级别胶质瘤(HGG)，使用MaZda软件提取T1WI对比增强(CE-T1WI)肿瘤纹理特征并分析比较两组间9个直方图参数，包括平均值(mean)、变异度(variance)、偏度(skewness)、峰度(kurtosis)和第1、10、50、90、99百分位数(Pere.1%、Pere.10%、Pere.50%、Pere.90%、Pere.99%)，采用多变量Logistic回归对有统计意义的参数进行建模并绘制受试者工作特征曲线(ROC)曲线评价在两组间差异有统计学意义的参数及多变量Logistic回归模型鉴别两者的效能。 结果LGG组的3个参数(kurtosis，pere.1%，pere.10%)高于HGG组(P均<0.05)，4个参数(mean，variance，pere.90，pere.99%)低于HGG组(P均<0.05)，而skewness、pere.50%这2个参数在两组间无明显差异(P均>0.05)；在两组间差异有统计学意义的7个参数中，variance鉴别两者的效能最佳，灵敏度、特异度及AUC分别为79.4%、86.2%、0.878，7个参数建立的多变量Logistic回归模型的效能优于所有参数，灵敏度、特异度及AUC分别为87.3%、79.3%、0.882。 结论基于CE-T1WI的直方图参数可有效鉴别LGG与HGG，而以两组间差异有统计学意义直方图参数建立的多变量Logistic回归模型诊断效能更佳。

Grading cerebral gliomas using texture analysis based on contrast-enhanced T1WI images

ObjectiveTo investigate the value of texture analysis based on contrast-enhanced T1-weighted images(CE-T1WI)in grading cerebral gliomas. MethodsFrom October 2014 to October 2019, 29 patients with low grade gliomas(LGG)and 63 patients with high grade gliomas(HGG), which were confirmed by postoperative pathology in Affiliated Hospital of Yangzhou University, were retrospectively reviewed.MaZda software was used to extract texture features of CE-T1WI of gliomas, and then 9 histogram parameters(mean, variance, skewness, kurtosis, Pere.1%, Pere.10%, Pere.50%, Pere.90%, Pere.99%)between the two groups were compared.Multivariate logistic regression model was established with the parameters which showed statistical differences between the two groups, and then receiver operating characteristic curve(ROC)was drew to evaluate the efficiency of the model. ResultsThree parameters(kurtosis, Pere.1%, Pere.10%)were higher in LGG than HGG(all P<0.05). Four parameters(mean, variance, Pere.90, Pere.99%)were lower in LGG than HGG(all P<0.05). There were no significant differences of two parameters(skewness, Pere.50%)between the two groups(both P>0.05). Among the parameters which showed statistically significant differences between the two groups, variance had the highest efficiency with sensitivity, specificity and area under the curve(AUC)of 79.4%, 86.2% and 0.878.The efficiency of the multivariate logistic regression model established with the seven parameters was better than that established with all the parameters, and its sensitivity, specificity and AUC were 87.3%, 79.3% and 0.882, respectively. ConclusionHistogram parameters based on CE-T1WI can effectively distinguish LGG from HGG, and the multivariate logistic regression model with histogram parameters which showed statistically significant difference between the two groups is more effective.