双指数及拉伸指数模型DWI在肝癌和局灶性结节增生鉴别中的联合应用

The combined application of DWI biexponential and stretched exponential model in the differentiation of hepatocellular carcinoma and focal nodular hyperplasia

目的探讨MRI体素内不相干运动(IVIM)弥散加权成像(DWI)双指数模型及拉伸指数模型多个参数在肝细胞癌(HCC)和肝脏局灶性结节增生(FNH)鉴别诊断中的联合应用价值。方法回顾分析本院经术后病理或穿刺活检确诊的26例HCC和17例FNH患者的MRI图像和IVIM-DWI后处理参数图。测量肿瘤病灶双指数模型参数:真实弥散系数(D)值、灌注相关弥散系数(D^*)值、灌注分数(f)值和拉伸指数模型参数:分布弥散系数(DDC)值、体素内弥散成分不均质性(α)值,并比较其在两组间的差异。使用受试者工作特征(ROC)曲线结合二项Logistic回归分析D、D^*、DDC、双指数模型参数联合(D+D^*)以及双指数拉伸指数模型参数联合(D+D^*+DDC)在组间的诊断效能,并使用MedCalc软件对曲线下面积(AUC)进行比较。结果HCC组病灶D、D^*和DDC值均低于FNH组(P均<0.05),f和值在两组间无统计学差异。D、D^*和DDC值鉴别两组的最佳阈值分别为1.165×10^-3mm^2/s、43.000×10^-3mm^2/s和1.450×10^-3mm^2/s,对应的约登指数分别为0.708、0.452和0.572,曲线下面积(AUC)分别为0.876、0.717和0.845。AUC从大到小排序为:AUCD+D^*+DDC>AUCD+D^*>AUCD>AUCDDC>AUCD^*。AUCD+D^*+DDC和AUCD+D^*无统计学差异(P=0.163)。AUCD和AUCDDC无统计学差异(P=0.273),但两者均明显>AUCD^*(P<0.001,P=0.015)。结论双指数和拉伸指数模型DWI参数联合(D+D^*+DDC)具有较高的诊断效能,可有效鉴别HCC和FNH,单独运用双指数模型(D+D^*)可达到近似两者联合的诊断效能。

Objective To explore the value of the combined application of multiple parameters in the biexponential and stretched exponential models of magnetic resonance(MRI)intravoxel incoherent motion(IVIM)diffusion weighted imaging(DWI)in the differential diagnosis of hepatocellular carcinoma(HCC)and focal nodular hyperplasia(FNH).Methods MRI and IVIM-DWI post-processing parameter images of 26 patients with HCC and 17 patients with FNH confirmed by postoperative pathology or biopsy in our hospital were retrospectively analyzed.Parameters of the biexponential model in terms of true diffusion coefficient(D)value,perfusion correlation diffusion coefficient(D^*)value,perfusion fraction(f)value and stretched exponential model parameters:distributed diffusion coefficient(DDC)value,diffusion heterogeneity in voxels value of tumor lesion were measured,and the differences of them between two groups were compared.The intergroup diagnostic efficacy of D,D^*,DDC,biexponential model parameters combination(D+D^*),biexponential and stretched exponential models parameters combination(D+D^*+DDC)was analyzed by using receiver operating characteristic(ROC)curve and binomial Logistic regression,and the area under the curve(AUC)was compared by MedCalc software.Results The D,D^*and DDC values of lesions in the HCC group were lower than those in the FNH group(all P<0.05),and the f and values were not statistically different between the two groups.The optimal thresholds of D,D^*and DDC values for identifying the two groups were 1.165×10^-3 mm^2/s,43.000×10^-3 mm^2/s and 1.450×10^-3 mm^2/s,respectively.The corresponding Youden index was 0.708,0.452 and 0.572,and the area under the curve(AUC)was 0.876,0.717 and 0.845,respectively.The descending order of AUC was AUCD+D^*+DDC>AUCD+D^*>AUCD>AUCDDC>AUCD^*.There were no statistically difference between AUCD+D^*+DDC,AUCD+D^*(P=0.163).AUCD and AUCDDC had no statistically difference(P=0.273),however,both of them were significantly higher than AUCD^*(P<0.001,P=0.015).Conclusion The combination of IVIM-DWI biexponential and stretched exponential models parameters(D+D^*+DDC)have a high diagnostic efficiency,and can effectively identify HCC and FNH.The application of the biexponential model(D+D^*)alone can achieve a similar diagnostic efficiency compared to the combination of the two models.