增强磁共振成像纹理参数对胶质母细胞瘤、原发性中枢神经系统淋巴瘤和单发转移瘤的鉴别诊断价值

目的探讨增强MRI纹理参数鉴别胶质母细胞瘤(GBM)、原发性中枢神经系统淋巴瘤(PCNSLs)及单发转移瘤(sMT)的可行性。方法回顾性分析2016年1月至2018年9月苏北人民医院收治的经手术或穿刺活检病理确诊的30例GBM、20例PCNSLs及21例sMT患者。应用Mazda软件勾画病灶感兴趣区并提取297个纹理参数,应用Fisher系数、分类错误联合平均相关系数提取(POE+ACC)、交互信息(MI)及三者联合FPM提取法筛选出30个特征参数。计算组内相关系数(ICC)评价观察者间测量纹理参数的一致性。比较GBM、PCNSLs和sMT纹理参数差异。将30个纹理参数纳入b11数据包进行原始数据分析(RDA)、主成分分析(PCA)、线性判别分析(LDA)、非线性判别分析(NDA)统计。结果两名观察者测量30个纹理参数的ICC为0.813~0.892,均≥0.75,表现出良好的一致性。30个纹理参数中,仅S(4,4)SumAverg、Teta3在GBM、sMT和PCNSLs这3种肿瘤间的差异无统计学意义(P均>0.05),其余参数在这3种肿瘤间的差异均有统计学意义(P均<0.05或0.01)。b11数据包RDA、PCA、LDA、NDA这4种分析法诊断正确例数分别为41例、40例、67例、62例,诊断准确率分别为57.8%、56.3%、94.4%、87.3%。其中,NDA分析法诊断准确率最高。结论增强MRI能够提取定量纹理参数,有助于鉴别脑胶质母细胞瘤、原发性淋巴瘤及单发转移瘤。

Value of magnetic resonance imaging texture parameters in differentiating brain glioblastoma,primary central nervous system lymphoma and single metastatic tumor

Objective To investigate the feasibility of magnetic resonance imaging(MRI)texture parameters in differentiating glioblastomas(GBM),primary cerebral lymphomas(PCNSLs)and single metastatic tumor(sMT).Methods A total of 30 cases of GBM,20 cases of PCNSLs and 21 cases of sMT pathologically confirmed by surgery or needle biopsy were analyzed retrospectively.Delineation of region of interest(ROI)and extraction of 297 texture parameters were performed by using Mazda software.Thirty texture parameters were selected by the joint extraction method of FPM(Fisher,POE+ACC and MI).Intraclass correlation coefficient(ICC)was analyzed to test the consistency of two observers.The texture parameters of GBM,PCNSLs and sMT were compared.Thirty texture parameters were included in b11 data packets for raw data analysis(RDA),principal component analysis(PCA),linear analysis(LDA)and nonlinear analysis(NDA).Results The ICC of 30 texture parameters measured by the two observers was 0.813-0.892,all≥0.75,showing good consistency.Among the 30 texture parameters,only S(4,4)SumAverg and Teta3 had no significant statistical difference(both P>0.05)but the other parameters had significant difference among the three kinds of tumors(all P<0.05 or 0.01).The correct diagnosis cases of RDA,PCA,LDA and NDA by b11 packet analysis were 41,40,67 and 62 respectively,and the correct rates were 57.8%,56.3%,94.4%and 87.3%,respectively.Among them,the correct rate of NDA analysis was the highest.Conclusion Enhanced MRI can extract quantitative texture parameters,which is helpful to distinguish GBM,PCNSLs and sMT.