基于MR组学特征的诺模图(Nomogram)在术前预测宫颈癌淋巴血管间隙侵犯中应用的初步研究  

A Preliminary Study on the Application of Nomogram Based on MR Histological Characteristics in Predicting Lymphatic Vascular Space Invasion of Cervical Cancer before Operation

作　　者：何月明[1] 陈思琳 马跃昆 罗锦文[2] 范良生[3] HE Yueming;CHEN Silin;MA Yuekun(The Fifth Affiliated Hospital of Guangzhou Medical University,Guangzhou,Guangdong Province 510170,P.R.China)

机构地区：[1]广州医科大学附属第五医院妇科,510700 [2]广州医科大学附属第五医院影像科,510700 [3]广州医科大学附属第一医院妇产科,510120

出　　处：《临床放射学杂志》2022年第8期1565-1574,共10页Journal of Clinical Radiology

基　　金：广东高校生物靶向诊治与康复重点实验室项目(编号:2021KSY009)。

摘　　要：目的探讨基于MR组学特征的诺模图(Nomogram)在术前预测宫颈癌淋巴血管间隙侵犯(LVSI)的价值。方法回顾性分析2015年2月至2019年12月共110例经手术病理证实的[国际妇产科协会(FIGO)分期Ⅰ~ⅡA期]的早期宫颈癌患者病理及MR影像资料,年龄36~75岁,平均(57±13)岁,按照7∶3的比例随机分为训练组(77例)和验证组(33例),根据病理分为LVSI阳性组和LVSI阴性组。在MR上测量得肿瘤的最大径、肌层浸润深度;使用MaZda 4.6软件对MR图像矢状位T_(2)WI抑脂序列的肿瘤影像提取794个3D影像组学特征,并用软件内自带的B11模块特征选择,得到对术后病理分组最优的30个组学特征。然后再对最优的30个组学特征进行LASSO回归分析筛选与宫颈癌LVSI最相关的组学特征,将最相关的组学特征纳入多因素Logistic回归分析中构建预测模型。绘制受试者工作特征(ROC)曲线分析用于评价预测模型Nomogram图的诊断性能,采用Delong检验比较不同组学特征之间的曲线下面积(AUC),绘制Nomogram图、决策曲线评估预测模型的临床应用价值。结果最终选择得到3个组学特征并建立的影像组学模型显示预测模型对宫颈癌LVSI阳性组和LVSI阴性组有良好的鉴别。S(0,5,5)Entropy、135°_LngREmph、Gr_NonZeros是LVSI独立预测指标。ROC曲线分析中,MR组学模型Nomogram图预测LVSI在训练组AUC及95%CI、敏感度、特异度为0.762(0.656~0.868)、0.881、0.5714,验证组中对应的AUC及95%CI,敏感度、特异度为0.711(0.534~0.888)、0.3889、1.00,且两者差异无统计学意义(P>0.05);决策曲线在组学特征训练组、验证组中净收益分别为14.40和13.6,组学联合临床特征训练组、验证组中净收益分别为16.1和15.8。结论基于MR的影像组学对术前预测宫颈癌LVSI状态具有较好的应用价值,可用于宫颈癌术前预测LVSI,决策曲线提示有较好的临床实用性。Objective To evaluate the value of nomogram(Nomogram)based on MR histological characteristics in predicting lymphatic vascular space invasion of cervical cancer before operation.Methods The pathological and MR data of 110 patients with early cervical cancer confirmed by operation and pathology(FIGO stageⅠ-ⅡA)in our hospital from February 2013 to December 2019 were analyzed retrospectively.The patients,aged 36 to 75(57±13)years,were randomly divided into training group(n=77)and verification group(n=33).According to pathology,they were divided into two groups:positive group and negative group.The maximum diameter of the tumor and the depth of myometrial invasion(proportion)were measured on MR,and 7943 D imaging features were extracted from the sagittal TWI fat compression sequence of MR images by MaZda 4.6 software,and the B11 module feature selection in the software was used to obtain the 30 best features for postoperative pathological grouping.Then LASSO regression analysis was used to screen the most relevant features related to lymphatic vascular space invasion of cervical cancer,and the most relevant features were included in multivariate Logistic regression analysis to build a prediction model.The ROC curve analysis of the subjects’working characteristics was used to evaluate the diagnostic performance of the nomogram of the predictive model.Delong test was used to compare the area under the curve between different grouping features(AUC),to draw the nomogram and the decision curve to evaluate the clinical application value of the predictive model.Results The final selection of three histological features and the establishment of the imaging model showed that the predictive model could well distinguish between the group of lymphatic vascular space invasion and the group of non-lymphatic vascular space invasion of cervical cancer.S(0,5,5)Entropy、135°_LngREmph、Gr_NonZerosare independent predictors of lymphatic vascular space invasion.In the ROC curve analysis,the MR group model nomogram predicted lym

关 键 词：磁共振成像 纹理分析 宫颈癌 淋巴血管间隙侵犯 

分 类 号：R445.2[医药卫生—影像医学与核医学] R737.33[医药卫生—诊断学]