联合临床病理参数及多序列MRI-Rad评分的影像组学列线图预测胰腺导管腺癌术后生存的价值  

The value of radiomics nomogram combing clinicopathological parameters and Rad-scores of multisequence MRI for predicting postoperative survival in pancreatic ductal adenocarcinoma

作　　者：韩伟光 苏水霞 王青 王铮[1] HAN Weiguang;SU Shuixia;WANG Qing;WANG Zheng(Department of Hepatobiliary Surgery,the First Affiliated Hospital of Xi'an Jiaotong University,Shaanxi Xi'an 710061,China;Department of General Surgery,the Second Affiliated Hospital,Air Force Medical University,Shaanxi Xi'an 710038,China.)

机构地区：[1]西安交通大学第一附属医院肝胆外科,陕西西安710061 [2]空军军医大学第二附属医院普通外科,陕西西安710038

出　　处：《现代肿瘤医学》2024年第13期2387-2393,共7页Journal of Modern Oncology

摘　　要：目的:探讨联合临床病理参数和多序列MRI-Rad评分的影像组学列线图在胰腺导管腺癌(pancreatic ductal adenocarcinoma,PDAC)患者术后生存预测中的价值。方法:回顾性分析我院2016年06月至2020年06月间行根治性切除手术并经病理确诊的197例PDAC患者胰腺MRI及临床病理资料。将患者按照7∶3比例划分为训练集(138例)与验证集(59例)。从T2加权成像(T2-weightedimaging,T2WI)、增强T1加权成像(T1-weightedimaging,T1WI)和扩散加权成像(diffusionweightedimaging,DWI)3个序列MRI图像中共提取360个影像组学特征并通过特征组间和组内相关系数(inter-and intra-class correlation coefficient,ICC)分析,保留ICC>0.75的特征。应用最小绝对收缩和选择算子(least absolute shrinkage and selection operator regression,LASSO)进一步筛选特征并建立MRI影像组学评分(radiomics score,Rad-score)。根据时间依赖性受试者工作特征(receiver operating characteristic,ROC)曲线下面积(area under curve,AUC)分析Rad评分预测1年和3年无病生存期(disease-free survival,DFS)和总生存期(overall survival,OS)的性能,并基于Rad评分截断值将患者分为高Rad评分组与低Rad评分组,使用Kaplan-Meier(K-M)生存分析评估Rad评分与DFS和OS关系。通过单变量和多变量Cox回归分析确定与DFS和OS相关的临床病理独立预测因素并构建临床病理模型。利用肿瘤T分期、N分期和M分期信息构建TNM分期系统模型。联合临床病理独立危险因素及Rad评分制定影像组学列线图(nomogram),利用Harrell's一致性指数(C-index,C指数)、校准曲线(calibration curves)和决策曲线分析(decision curve analysis,DCA)评估列线图的效能并在3种模型间进行比较。结果:通过ICC和LASSO分析,从多序列MRI中共筛选7个影像组学特征并建立Rad评分。Rad评分预测1年和3年DFS和OS的AUC分别为0.807、0.795和0.839、0.812,低Rad评分组患者DFS和OS显著优于高Rad评分组(P<0.001)。多因素Cox回归Objective:To explore the value of radiomics nomogram combining clinicopathological parameters and Rad-score of multisequence MRI for predicting postoperative survival in patients with pancreatic ductal adenocarcinoma (PDAC).Methods:Clinicopathological data and preoperative MRI of 197 patients with PDAC who underwent radical resection between June 2016 and June 2020 confirmed in our hospital were retrospectively analyzed.The patients were divided into a training set(138 cases) and a validation set(59 cases) according to a 7∶ 3 ratio.A total of 360 radiomics features were extracted from T2-weightedimaging(T2WI),enhanced T1-weightedimaging(T1WI),and diffusion weightedimaging(DWI) of multisequence MRI,and analyzed using inter-and intra-class correlation coefficients(ICC) to retain the features with ICC > 0.75.The least absolute shrinkage and selection operator regression(LASSO) was applied to further select the features and establish the MRI radiomics score(Rad-score).The performance of the Rad-score in predicting the performance of 1-and 3-year disease-free survival(DFS) and overall survival(OS) was evaluated using the area under curve(AUC) of the time-dependent receiver operating characteristic(ROC).Based on the cutoff value of the Rad-score,the patients were categorized into high Rad-score group and low Rad-score group,and the relationship between Rad-score and DFS and OS was assessed using Kaplan-Meier(K-M) survival analysis.Univariate and multivariate Cox regression were performed to identify clinicopathological independent risk factors associated with DFS and OS and to construct clinicopathological model.TNM model was constructed using the T staging,N staging,and M staging data.Radiomics nomograms were developed by combining clinicopathological independent risk factors and Rad-scores,and the effectiveness was evaluated using Harrell's consistency index(C-index),calibration curves,and decision curve analysis(DCA).Then the three models were compared.Results:A total of seven radiomics features were selected from

关 键 词：胰腺导管腺癌 磁共振成像 影像组学 列线图 生存 

分 类 号：R735.9[医药卫生—肿瘤]