MRI纹理分析预测胶质母细胞瘤患者MGMT蛋白表达的研究  

A study on the prediction of MGMT protein expression in patients with glioblastoma through MRI texture analysis

作　　者：李梦双 刘耀赛 董丽娜 罗涛 徐凯 LI Mengshuang;LIU Yaosai;DONG Lina;LUO Tao;XU Kai(Department of Medical Imaging,the Affiliated Hospital of Xuzhou Medical University,Xuzhou 221006;Department of Neurosurgery,the Affiliated Hospital of Xuzhou Medical University,Xuzhou 221006)

机构地区：[1]徐州医科大学附属医院医学影像科,江苏徐州221006 [2]徐州医科大学附属医院神经外科,江苏徐州221006

出　　处：《实用放射学杂志》2021年第1期9-12,共4页Journal of Practical Radiology

基　　金：江苏省研究生实践创新计划项目(SJCX19_0940).

摘　　要：目的探讨MRI纹理分析在预测胶质母细胞瘤(GBM)患者MGMT蛋白表达状态的应用价值.方法回顾性选取经手术病理证实的123例GBM患者,对其T1WI增强图像进行纹理分析,获取整个病灶的偏度、峰度、熵值、能量、平均数、中位数及标准差7个定量参数,进行统计学分析.结果纹理分析获得的定量参数中,峰度、平均数、中位数、标准差4个参数在MGMT(+)组与MGMT(-)组间差异无统计学意义,偏度、熵值和能量3个参数2组间差异有统计学意义(P<0.001).偏度、熵值和能量3个定量参数在诊断敏感性、特异性和准确性方面各具优势;对偏度、熵值和能量进行多参数联合分析,曲线下面积(AUC)值为0.972,当阈值为0.918时,其诊断敏感性为95.0%、特异性为96.8%、准确性为95.8%,明显高于利用单个纹理分析定量参数的预测效能.结论MRI纹理分析部分定量参数(偏度、熵值和能量)有助于预测GBM患者MGMT蛋白表达状态,指导临床个性化治疗;多参数联合分析预测效能更高.Objective To explore the application value of MRI texture analysis on predicting the MGMT protein expression in patients with glioblastoma(GBM).Methods 123 cases of GBM confirmed by operative pathology were retrospectively collcctcd/Tcxture analysis on their T WI enhanced images was conducted to acquire seven quantitative parameters including the skewness,kurtosis,entropy,energy,mean,median and standard deviation of the whole lesion,and these data were statistically analyzed.Results Among the quantitative parameters obtained,there was no significant difference between MGMT(+)group and MGMT(-)group on the parameters of kurtosis,mean,median and standard deviation,while significant differences were observed on the other three parameters of skewness,entropy and energy(P〈0.001).Skcwncss,cntropy and energy had their own advantages on the scnsitivity,spccificity and accuracy of diagnosis.When multi-parameters of skewness,entropy and energy were analyzed,the value of AUC was 0.972/The diagnostic sensitivity,specificity and accuracy were 95.0%,96.8%,95.8%respectively if the threshold was taken as 0.918,which was significantly higher than that by using single quantitative parameter in the prediction efficiency of texture analysis.Conclusion Some specific quantitative parameters(skewness,entropy and energy)of MRI texture analysis are beneficial to the prediction of the MGMT protein expression of GBM and so as to guide clinical individualized treatment,and the combination analysis of multi-parameters has higher prediction efficiency.

关 键 词：纹理分析 磁共振成像 胶质母细胞瘤 MGMT 

分 类 号：R445.2[医药卫生—影像医学与核医学] R739.41[医药卫生—诊断学]