MR全容积T2WI信号直方图术前预测子宫内膜癌病理分级的初步研究  

Predicting tumor grade in endometrial carcinoma based on whole tumor T2WI histogram analysis:A preliminary study

作　　者：张宇辰[1] 邓颖 敦旺欢[3] 王茵 闫斌 ZHANG Yuchen;DENG Ying;DUN Wanghuan;WANG Yin;YAN Bin(Department of Nuclear Medicine,the First Affiliated Hospital of Xi'an Jiaotong University,Shaanxi Xi'an 710061,China;Department of Rehabilitation Medicine,the First Affiliated Hospital of Xi'an Jiaotong University,Shaanxi Xi'an 710061,China;Department of Medical Imaging,the First Affiliated Hospital of Xi'an Jiaotong University,Shaanxi Xi'an 710061,China;Department of Radiology,Shaanxi Provincial Cancer Hosptial,Shaanxi Xi'an 710061,China)

机构地区：[1]西安交通大学第一附属医院核医学科,陕西西安710061 [2]西安交通大学第一附属医院康复医学科,陕西西安710061 [3]西安交通大学第一附属医院影像科,陕西西安710061 [4]陕西省肿瘤医院放射科,陕西西安710061

出　　处：《现代肿瘤医学》2022年第22期4150-4155,共6页Journal of Modern Oncology

基　　金：国家自然科学基金(编号:81871331);陕西省西安市创新能力强基计划-医学研究项目(编号:21YXYJ0102)。

摘　　要：目的:探讨肿瘤全域T2WI直方图分析术前预测子宫内膜癌(endometrial carcinoma,EC)病理分级的应用价值。方法:回顾性收集经术后病理证实为EC患者111例,其中高级别EC患者34例,低级别组77例。所有患者术前均接受3.0-T MR检查。采用3D Slicer软件,于轴位T2WI上,在包含肿瘤实质的每层图像上沿肿瘤边缘手动勾画感兴趣区(region of interest,ROI),将其累加,获得三维ROI的信号强度直方图及其参数(包括最大值、最小值、平均值、偏度、峰度、第5、10、25、50、75、90、95百分位数)。组内相关系数(intraclass correlation coefficient,ICC)用于评价T2WI直方图不同观测者间测量的变异性。比较高级别与低级别EC组间各参数的差异,绘制受试者工作特性(receiver operating characteristic,ROC)曲线,评估其对EC病理分级预测的效能。结果:T2WI直方图参数中,仅峰度值在高级别与低级别组间具有统计学差异(P<0.010),高级别EC具有更高的峰度(2.049 vs 1.893),ROC曲线下面积为0.654;阈值取0.813,预测高级别EC的敏感性、特异性及准确性分别为79.4%、58.4%及64.9%。所有直方图参数均显示出一般到较好的观测者间一致性(ICC=0.731~0.973)。结论:肿瘤全域T2WI直方图分析有助于术前预测EC病理级别,峰度可能是唯一有价值的预测指标。Objective:To explore the application value of whole tumor T2WI histogram analysis in predicting the tumor grade of endometrial carcinoma(EC)before operation.Methods:111 patients with EC confirmed by postoperative histopathology were collected retrospectively,including 34 patients with high-grade EC and 77 patients in low-grade group.All patients underwent 3.0-T MR examination before operation.Using 3D slicer software,manually draw ROI along the tumor edge on each layer of image containing tumor margin on axial T2WI,and accumulate them to obtain the signal intensity histogram of 3D ROI and its parameters(including the maximum value,minimum value,average value,skewness,kurtosis,5th,10th,25th,50th,75th,90th and 95th percentile values).The interobserver correlation coefficient(ICC)was used to evaluate the variability of T2WI histogram measurement.The differences of parameters between high-and low-grade EC were compared,and the receiver-operating characteristic(ROC)curve was drawn for the different parameters to evaluate the efficiency of graded diagnosis of EC.Results:Among the histogram parameters for distinguishing high-and low-grade EC,only the kurtosis value had statistical difference between the two groups(P<0.010),high-grade EC had higher kurtosis value(2.049 vs 1.893),the area under ROC curve was 0.654.A kurtosis value>0.813 predicted high-grade EC with a sensitivity,specificity and accuracy of 79.4%,58.4%and64.9%respectively.All histogram parameters showed good to excellent inter observer reliability(ICC=0.731~0.973).Conclusion:The whole tumor T2WI histogram analysis is helpful to predict the tumor grade of EC before operation,and kurtosis may be the only valuable predictor.

关 键 词：子宫内膜癌 磁共振 T2加权成像 直方图分析 病理分级 

分 类 号：R737.33[医药卫生—肿瘤]