DCE-MRI全定量参数及纹理分析应用于肝脏良恶性病变的诊断效能

发布时间：2018-03-20 01:00

  本文选题：肝脏肿瘤　切入点：磁共振成像　出处：《重庆医科大学》2017年硕士论文　论文类型：学位论文

【摘要】：目的:探讨动态增强磁共振成像(Dynamic contrast-enhanced magnetic resonance imaging,DCE-MRI)全定量参数及Ktrans纹理特征在鉴别肝脏良恶性病变中的诊断效能。方法:回顾性分析60例经病理或临床证实患有肝脏占位性病变患者的DCE-MRI图像资料。原始图像资料导入后处理工作站,应用双输入ETK(Extended Tofts-Kermode,ETK)模型计算全定量参数容积转运常数(Volume transfer constant,Ktrans)、血管外细胞外容积分数(Volume fraction of extravascular extracellular space,Ve)、速率常数(Interstitium-to-plasma rate constant,Kep)、血浆容积分数(volume fraction of plasma,Vp)。应用灰度直方图(Gray-level Histogram)纹理分析提取Ktrans图像的纹理特征,如能量(Energy)、熵(Entropy)、偏度系数(Skewness)、峰度系数(Kurtosis)、均匀度(Uniformity)。将所获得的全定量参数及Ktrans灰度直方图纹理特征通过独立样本t检验或Mann-Whitney U检验进行统计学分析。并通过二元Logistic回归分别将两组影像学参数拟合为两两联合诊断模型,包括Ktrans-Kep、Ktrans-Ve、Kep-Ve和Energy-Entropy、Energy-Skewness、Energy-Kurtosis、Entropy-Skewness、Entropy-Kurtosis、Skewness-Kurtosis。计算各影像学指标及对应联合诊断模型对病灶的预测频率。以病理或临床诊断为金标准,利用频率变量绘制受试者工作特征(Receiver operating characteristic,ROC)曲线并计算曲线下面积(Area under the curve,AUC)。利用最大约登指数分别探讨两组影像学指标及对应两两联合诊断模型的敏感度、特异度。结果:全定量参数中Ktrans、Kep、Vp符合正态分布及方差齐同性,Ve符合正态分布,不符合方差齐同性。恶性病灶组Ktrans及Kep(1.233±0.613,1.777±0.848)明显高于良性病灶组(0.532±0.371,0.942±1.151),组间差异具有统计学意义(独立样本t检验,t值分别为�c5.064和�c3.004,P值分别为0.000和0.004)。良性病灶组Ve(4.343±2.996)明显高于恶性病灶组(1.271±0.675),组间差异具有统计学意义(Mann-Whitney U检验,P值为0.000)。恶性病灶组Vp(0.332±0.243)高于良性病灶组(0.219±0.267),组间差异不显著,无统计学意义(独立样本t检验,t值为�c1.601,P值为0.116)。全定量参数及两两联合诊断模型ROC曲线下面积分别为0.841(Ktrans)、0.848(Ve)、0.815(Kep)、0.844(Ktrans-Kep)、0.929(Ktrans-Ve)、0.808(Kep-Ve),根据最大约登指数分析,Ktrans-Ve诊断效能最佳,特异度最高(100%);Ve诊断效能次于Ktrans-Ve,敏感度最高(85.7%)。Ktrans直方图纹理特征均符合正态分布及方差齐同性。良性病灶组Energy、Kurtosis(0.762±0.264,2.100±0.894)高于恶性病灶组(0.610±0.265,1.126±0.751),组间差异具有统计学意义(独立样本t检验,t值分别为�c2.022、�c4.187,P值分别为0.049、0.000)。恶性病灶组Entropy、Skewness(1.056±0.683、0.489±0.579)高于良性病灶组(0.658±0.695、0.244±0.583),组间差异具有统计学意义(独立样本t检验,t值分别为2.033、4.450,P值分别为0.048、0.000)。恶性病灶组Uniformity(�c0.543±0.432)高于良性病灶组(�c0.555±0.607),组间差异无统计学意义(独立样本t检验,t值为0.067,P值为0.947)。纹理特征及两两联合诊断模型ROC曲线下面积分别为0.654(Energy)、0.645(Entropy)、0.825(Skewness)、0.828(Kurtosis)、0.643(Energy-Entropy)、0.857(Energy-Skewness)、0.842(Energy-Kurtosis)、0.835(Entropy-Skewness)、0.842(Entropy-Kurtosis)、0.839(Skewness-Kurtosis)。Energy、Entropy、Energy-Entropy三组ROC曲线不具有统计学意义。Energy-Skewness模型诊断效能最佳,曲线下面积为0.857。根据最大约登指数获得敏感度及特异度,Skewness、Energy-Skewness、Entropy-Skewness、Entropy-Kurtosis特异度高(81.5%),Kurtosis、Energy-Skewness、Energy-Kurtosis、Entropy-Skewness敏感度高(87.0%)。结论:DCE-MRI获得的全定量参数Ktrans、Kep、Ve及Ktrans灰度直方图纹理特征Energy、Entropy、Skewness、Kurtosis能够有效鉴别肝脏良恶性病灶。而全定量参数Vp及纹理特征Uniformity不能有效鉴别肝脏良恶性病灶。全定量参数两两联合诊断模型Ktrans-Ve,纹理特征两两联合诊断模型Energy-Skewness在各组内分析中具有更高诊断效能,对鉴别诊断肝脏良恶性病变方面较其他影像学指标及联合诊断模型更具优势。
[Abstract]:Objective: To investigate the dynamic enhanced magnetic resonance imaging (Dynamic contrast-enhanced magnetic resonance imaging, DCE-MRI) the diagnostic efficacy of quantitative parameters and Ktrans texture features in the differential diagnosis of benign and malignant liver lesions. Methods: a retrospective analysis of 60 cases proved by pathology or clinical withLiver accounted for DCE-MRI image data of lesions. Into the original image data the workstation, using double input ETK (Extended Tofts-Kermode ETK) model to calculate the total volume transport quantitative parameters (Volume transfer constant, Ktrans constant), extravascular extracellular volume fraction (Volume fraction of extravascular extracellular space, Ve), the rate constant (Interstitium-to-plasma rate constant Kep (volume), plasma volume fraction fraction of plasma, Vp). The application of gray histogram (Gray-level Histogram) extract texture feature texture analysis Ktrans image, such as Volume (Energy), entropy (Entropy), skewness, kurtosis coefficient (Skewness) (Kurtosis), evenness (Uniformity). The quantitative parameters of gray histogram and Ktrans texture features obtained were analyzed by independent samples t test or Mann-Whitney U test. Logistic regression and by two yuan respectively two imaging parameters fitting 22 combined diagnosis model, including Ktrans-Kep, Ktrans-Ve, Kep-Ve and Energy-Entropy, Energy-Skewness, Energy-Kurtosis, Entropy-Skewness, Entropy-Kurtosis, Skewness-Kurtosis. to calculate the combined imaging diagnosis index and the corresponding model for the prediction of lesion frequency. The pathological or clinical diagnosis as the gold standard, drawing the subjects using the frequency variable operating characteristic (Receiver operating characteristic, ROC) curve of area under the curve was calculated (Area under the curve, AUC) were discussed using the maximum Youden index. Sensitivity of the two groups of combined imaging diagnosis index model and the corresponding 22 specificity. Results: Ktrans, the quantitative parameters of Kep and Vp with normal distribution and homogeneity of variance, Ve with normal distribution, does not meet the variance homogeneity. Malignant lesions group Ktrans and Kep (1.233 + 0.613,1.777 + 0.848) significantly higher than that in benign lesion group (0.532 + 0.371,0.942 + 1.151), a statistically significant difference between the groups (independent sample t test, the T values were 5.064 and 3.004??, P = 0 and 0.004). Benign lesions group Ve (4.343 + 2.996) was significantly higher than that of malignant disease lesion group (1.271 + 0.675). The differences between groups was statistically significant (Mann-Whitney U test, P = 0). Malignant lesions group Vp (0.332 + 0.243) higher than that of benign lesions group (0.219 + 0.267), no significant difference between the groups, no statistical significance (independent samples t test, t =? 1.601, P = 0.116). The quantitative parameters and 22 combined diagnosis The area of fault model under the ROC curve were 0.841 (Ktrans), 0.848 (Ve), 0.815 (Kep), 0.844 (Ktrans-Kep), 0.929 (Ktrans-Ve), 0.808 (Kep-Ve), according to the analysis of the maximum Youden index, Ktrans-Ve diagnosis efficiency was the best, the highest specificity (100%); Ve diagnostic efficacy after Ktrans-Ve, sensitivity the highest (85.7%).Ktrans histogram texture features accord with normal distribution and homogeneity of variance. Benign lesions group Energy, Kurtosis (0.762 + 0.264,2.100 + 0.894) higher than that of malignant lesions group (0.610 + 0.265,1.126 + 0.751), a statistically significant difference between the groups (independent sample t test, t =? 2.022, 4.187?, P = 0.049,0.000). Malignant lesions group Entropy, Skewness (1.056 + 0.683,0.489 + 0.579) higher than that of benign lesions group (0.658 + 0.695,0.244 + 0.583), a statistically significant difference between the groups (independent sample t test, t = 2.033,4.450, P = 0.048,0.000) evil. 鎬х梾鐏剁粍Uniformity(?0.543卤0.432)楂樹簬鑹�妫�

本文编号：1636781