常规MRI分形分析鉴别胶质母细胞瘤与孤立性脑转移瘤

Conventional MRI fractal analysis in differentiating glioblastoma from solitary metastatic brain tumor

目的:探讨常规MRI分形分析鉴别脑胶质母细胞瘤(GBM)与孤立性脑转移瘤(sMT)的价值。方法:回顾性分析2015年11月-2022年8月于空军军医大学第二附属医院经手术病理证实的92例GBM及69例sMT的常规MRI平扫及增强影像。两名放射科医生基于增强后T1WI(T1CE)进行不同病变亚区的分割,包括强化区(CE)、非强化区(NCE)及水肿区(ED)。利用Image J及Fraclac插件进行分形分析,提取不同亚区分形维数(FD)及空隙度(LAC)参数。采用独立样本t检验评估两组间分形分析参数差异。二元Logistic回归分析用于多参数联合(Combination)模型构建。采用受试者操作特征(ROC)曲线及Hosmer-Lemeshow拟合优度检验评价参数/模型诊断效能及拟合优度。采用DeLong检验比较不同参数/模型的曲线下面积(AUC)。结果:GBM强化区FD(FDce,1.596±0.067)、强化区LAC(LACce,0.510±0.098)及水肿区FD(FDed,1.570±0.069)平均值高于sMT组(FDce=1.396±0.116;LACce=0.416±0.075;FDed=1.369±0.154),差异具有统计学意义(t=12.8、6.936、10.027,P均<0.001)。Combination模型诊断效能(AUC=0.964)优于最佳单参数FDce(AUC=0.928),差异具有统计学意义(P=0.0096)。Hosmer-Lemeshow拟合优度检验显示各参数/模型具有良好的区分能力(P均>0.1)。结论:常规MRI分形分析有助于鉴别GBM与sMT,Combination模型鉴别两者效能最佳,为临床决策提供理论依据。

Objective:To investigate the value of conventional MRI fractal analysis in differentiating glioblastoma(GBM)from solitary metastatic brain tumor(sMT).Methods:From November 2015 to August 2022,The MRI images of 92 cases of GBM and 69 cases of sMT confirmed by pathology were retrospectively analyzed in Second Affiliated Hospital of Air Force Military Medical University.Two radiologists segmented the lesions into different subregions based on enhanced T 1WI(T 1CE),including contrast enhanced region(CE),non contrast-enhanced region(NCE)and edema region(ED).Image J software and Fraclac plugin were used for fractal analysis,and fractal dimension(FD)and lacunarity(LAC)of different subregions were extracted.Independent sample t-test was used to analyze the differences of fractal analysis parameters between the two groups.Multi-parameter combination model(Combination)was constructed by binary logistic regression.The diagnostic efficacy and goodness of fit of the parameters/model were evaluated by receiver operating characteristic(ROC)curves and Hosmer-Lemeshow tests.Differences between the AUCs were compared by using a Delong test.Results:The mean values of FD in the CE(FDce,1.596±0.067),LAC in the CE(LACce,0.510±0.098),FD in the ED(FDed,1.570±0.069)of GBM were higher than those in the sMT group(FDce=1.396±0.116;LACce=0.416±0.075;FDed=1.369±0.154),and differences were statistically significant(t=12.8,6.936,10.027,P<0.001).The diagnostic efficacy of the Combination model(AUC=0.964)was higher than that of optimal single parameter FDce(AUC=0.928),with a statistically significant difference(P=0.0096).The Hosmer Lemeshow test showed that all parameters/model had good discrimination ability(all P>0.1).Conclusion:The conventional MRI fractal analysis can be applied in differentiating GBM from sMT,and the Combination model has the best diagnostic efficacy,providing an important reference for clinical decision-making.