MRI影像学组预测宫颈鳞癌盆腔淋巴结转移的价值  

Value of MRI-based radiomics for the preoperative prediction of pelvic lymph node metastasis of cervical squamous cell carcinoma

作　　者：王岩[1] 赵青 孙东兰[1] 裴晓娇 陈晓[1] WANG Yan;ZHAO Qing;SUN Donglan;PEI Xiaoj iao;CHEN Xiao(Department of Obstetrics and Gynecology,the Fourth Hospital of Shijiazhuang,Shijiazhuang 050000,China)

机构地区：[1]石家庄市第四医院妇产科,河北石家庄050000

出　　处：《实用放射学杂志》2022年第2期273-277,306,共6页Journal of Practical Radiology

摘　　要：目的探讨MRI影像学组预测宫颈鳞癌盆腔淋巴结(PLN)转移的价值.方法回顾性分析114例宫颈鳞癌患者资料.采用完全随机法按照7︰3比例将患者分为训练集(79例)和验证集(35例),根据手术病理结果分为PLN阳性组(27例)和PLN阴性组(87例).由2名具有3年及以上盆腔疾病诊断经验的影像科医师勾画感兴趣区(ROI),计算组内相关系数(ICC)分析一致性.基于扩散加权成像(DWI)和T2 WI序列进行特征选择与降维,并采用Logistic多元回归相关性分析筛选出有预测价值的影像组学特征.分析患者MRI图像特点,比较训练集和验证集患者临床特征.对患者MRI影像组学评分与PLN转移进行分层分析.绘制受试者工作特征(ROC)曲线分析MRI影像组学特征对PLN转移的预测价值.结果训练集中,与PLN阴性患者比较,PLN阳性患者MRI评估PLN阳性率高(P<0.05);验证集中,与PLN阴性患者比较,PLN阳性患者MRI测量肿瘤最大径>4 cm比例高,MRI评估PLN阳性率高(P<0.05);本研究共获得训练集患者2789个影像组学特征,基于DWI和T2 WI序列特征选择与降维,最终获得6个影像组学特征;6个影像组学特征参数均与PLN转移呈显著相关(P<0.05);PLN阳性和PLN阴性患者MRI影像组学评分在年龄、MRI测量肿瘤最大径、国际妇产科联盟(FIGO)分期、病理类型、分化程度、浸润程度、脉管癌栓和MRI评估PLN转移亚组中,差异有统计学意义(P<0.05);训练集中,各MRI影像组学特征参数预测PLN转移曲线下面积(AUC)分别为0.747、0.745、0.692、0.699、0.556、0.670,6个影像组学特征联合预测PLN转移的AUC为0.833,高于单个特征预测PLN转移效能,且以上均在验证组中验证.结论基于DWI和T2 WI序列建立的MRI影像学组模型判断宫颈鳞癌患者PLN转移灵敏度和特异度较高,可为临床治疗和预后判断提供较可靠的影像学依据.Objective To investigate the value of MRI-based radiomics for the preoperative prediction of pelvic lymph node(PLN)metastasis of cervical squamous cell carcinoma.Methods The data of 114 patients with cervical squamous cell carcinoma were analyzed retrospectively.Patients were divided into a training set(n=79)and a validation set(n=35)via the completely stochastic method at a ratio of 7︰3,and was divided into PLN positive(n=27)and PLN negative(n=87)group according to outcomes of operation and pathology.The region of interest(ROI)was delineated manually by two experienced radiologists in MR imaging,and the intraclass correlation coefficient(ICC)was used to evaluate the repeatability and consistency.Imaging features were extracted from diffusion weighted imaging(DWI)and T2 WI,and the feature selection and radiomics signature building were performed through multivariate Logistic regression method.The MRI image characteristics were analyzed,and the clinical characteristics of training set and validation set were compared.Stratified analyses were also performed.The receiver operating characteristic(ROC)curve was drawn to analyze the predictive value of MRI imaging features for the metastasis of PLN.Results Within the training set,the positive rate of PLN was higher in PLN positive patients than in PLN negative patients(P<0.05).Within the validation set,the proportion of tumor diameter>4 cm and the positive rate of PLN was higher in PLN positive patients than in PLN negative patients(P<0.05).A summary of 2789 imaging features were extracted from radiomics,and 6 PLN status-related features were extracted from the combination of DWI and T2 WI.Six features parameters were all significantly correlated with PLN metastasis(P<0.05).The MRI imaging scores of PLN positive and PLN negative patients had statistically significant differences in age,maximum diameter of tumor,International Federation of Gynecology and Obstetrics(FIGO)stage,pathological type,differentiation degree,invasion degree,vascular tumor thrombus and PLN meta

关 键 词：宫颈鳞癌 盆腔淋巴结转移 磁共振成像 影像学组 

分 类 号：R737.33[医药卫生—肿瘤] R445.2[医药卫生—临床医学]