ADC值联合纹理分析术前预测子宫内膜癌病理分级、肌层侵犯深度和淋巴血管间隙浸润的价值

Combination Analysis of ADC Value with Texture Analysis Based on T_(2)WI and CE-T_(1)WI for the Preoperative Prediction about Grade,DMI,LVSI in Endometrial Carcinoma

目的 探讨表观扩散系数(ADC)联合纹理分析术前预测子宫内膜癌(EC)的病理分级、肌层侵犯深度(DMI)和淋巴血管间隙浸润(LVSI)的价值.方法 回顾性分析134例患者经中国科学技术大学附属第一医院西区术后病理证实为EC的临床病理资料,测量所有患者的ADC值及T2WI和增强T1WI(CE-T1WI)序列的三维纹理特征.用独立样本t检验和Mann-Whitney U检验比较ADC值和纹理参数在不同组内的差异,将单因素分析有统计学意义的参数进行多因素Logistic回归分析并分别建立ADC值联合多序列纹理预测病理分级、DMI及LVSI的模型.用受试者工作特征曲线(ROC)评估各参数及模型的诊断效能.结果 ADC值在病理分级、DMI、LVSI组内均有统计学意义,曲线下面积(AUC)分别为0.762、0.743、0.764.对T2WI和CE-T1WI各提取的1037个纹理参数,经多种统计学方法多重降维,筛选出18个纹理参数.其中,original_shape_Sphericity、wavelet_HHH_gldm_Large De-pendence Emphasis(LDE)﹑wavelet_LLL_glcm_Inverse Difference Normalized(IDN)分别在病理分级、DMI、LVSI 组中AUC最大,分别为0.701、0.697、0.783.ADC值联合T2WI、CE-T1WI纹理特征术前预测病理分级、DMI、LVSI的AUC分别为0.876﹑0.866、0.889.结论 ADC值和多序列三维纹理特征在术前预测EC病理分级、DMI及LVSI中有较高的应用价值,且联合预测更为精准.

Objective To investigate the value of apparent diffusion coefficient(ADC)value combined with texture analysis(TA)in predicting grade,the depth of myometrial invasion(DMI),lymph vascular space invasion(LVSI)in patients with endometrial carcinoma(EC)preoperatively.Methods A retrospective analysis of 134 cases with EC confirmed by surgical pathology were enrolled in West Branch of the First Affiliated Hospital of University of Science and Technology of China.All patients underwent MRI examination and ADC of the lesions was measured.Additionally,the TA was performed on T_(2)WI and CE-T_(1)WI images to obtain the 3D texture parameters.The statistical analysis was conducted using independent sample t test and Mann-Whitney U test to compare the ADC values and texture parameters among different groups.Multivariate Logistic regression and the models of ADC values combined with multi-sequence texture to predict pathological grade,DMI,LVSI were given rise to the parameters which existed statistically essential differences from univariate analysis,and the diagnostic efficiency of each parameter and model were evaluated by drawing the receiver operating characteristic curve(ROC).Results The ADC values exerted an essential influence on different grade,DMI,LVSI,and the AUC were 0.762,0.743,0.764,respectively.By means of multiple dimensional reduction methods,18 texture parameters with high statistical significances were selected totally from 1037 extracted texture parameters,which extracted from T_(2)WI and CE-T_(1)WI sequence respectively.Among these texture parameters,the AUC of original_shape_Sphericity,wavelet_HHH_gldm_Large Dependence Emphasis(LDE),wavelet_LLL_glcm_Inverse Difference Normalized(IDN)were largest in evaluating grade,DMI,LVSI,which were 0.701,0.697,and 0.783,respectively.Besides,The combined prediction efficacy of ADC value with radiomic features in different grade,DMI,LVSI were 0.876,0.866,0.889,respectively.Conclusion ADC values and multi-sequence 3D TA express essential application value in preoperative prediction of pathological grade,DMI and LVSI in EC,which have better predictive performance when combined.