基于T2WI图像的影像组学列线图预测直肠癌同步肝转移的价值  

A radiomic nomogram based on T2WI for predicting synchronous liver metastasis of rectal cancer

作　　者：舒震宇[1] 方松华[1] 邵园[1] 毛德旺[1] 柴瑞[2] 陈愿君 龚向阳[1] Shu Zhenyu;Fang Songhua;Shao Yuan;Mao Dewang;Chai Rui;Chen Yuanjun;Gong Xiangyang(Department of Radiology,Zhejiang Province People′s Hospital,People's Hospital of Hangzhou Medical College,Hangzhou 310014,China;Department of Colorectal Surgery,Zhejiang Province People′s Hospital,People's Hospital of Hangzhou Medical College,Hangzhou 310014,China;GE China Medical Life Science Department Core Image Advanced Application Team,Shanghai 201210,China)

机构地区：[1]浙江省人民医院杭州医学院附属人民医院放射科,杭州310014 [2]浙江省人民医院杭州医学院附属人民医院肛肠外科,杭州310014 [3]GE中国医疗生命科学部核心影像高级应用团队,上海201210

出　　处：《中华放射学杂志》2019年第3期205-211,共7页Chinese Journal of Radiology

基　　金：浙江省卫生健康委员会面上项目(2019318944).

摘　　要：目的探讨基于T2WI图像的影像组学列线图预测直肠癌同步肝转移的价值。方法回顾性分析2012年4月至2018年5月浙江省人民医院收治的261例原发性直肠癌患者的影像和临床资料,其中101例伴有同步肝转移,将患者分为为训练组(n=182)和验证组(n=79)。选取每例患者的T2WI图像通过AK分析软件提取纹理特征,然后对训练组使用最小绝对收缩与选择算子算法(LASSO)进行降维后建立影像组学标签。使用单因素逻辑斯回归筛选独立临床危险因素并使用多变量逻辑斯回归结合影像组学标签构建预测模型并制作列线图。使用ROC评估列线图和影像组学标签在训练组中的准确性并通过验证组进行验证,最后基于列线图计算每例患者的同步肝转移风险系数评估其临床效能。结果从T2WI图像中共提取了328个纹理特征,使用LASSO算法降维后筛选出7个价值较大的特征,其中共生矩阵3个、游程矩阵4个。使用多因素逻辑斯回归构建包含了MRI肿瘤T分期和影像组学标签的预测模型并制作列线图,列线图预测同步肝转移的准确率在训练组和验证组中分别为0.862和0.844,评估列线图、影像组学标签和肿瘤T分期在所有患者中的准确性分别为0.857、0.832和0.663。基于列线图区分的高风险组与低风险组中的同步肝转移患者数差异无统计学意义(P>0.05)。结论基于T2WI图像的影像组学列线图可作为一种量化工具预测直肠癌同步肝转移。Objective To explore the clinical feasibility of predicting synchronous liver metastases based on MRI radiomics nomogram based on T2WI in rectal cancer.Methods The imaging and clinical data of 261 patients with primary rectal cancer admitted to Zhejiang People′s Hospital from April 2012 to May 2018 were retrospectively analyzed.101 patients were accompanied by synchronous liver metastasis All cases were divided into training group(n=182)and verification group(n=79).T2WI image of each patient was selected to extract texture features by AK analysis software of GE company.A radiomics signature was constructed after reduction of dimension in training group by the least absolute shrinkage and selection operator(LASSO).Univariate logistic regression was used to select for independent clinical risk factors and multivariate logistic regression along with imaging omics tags were used to construct predictive models and nomogram.ROC was used to assess the accuracy of the nomogram in the training group and to verify them by the validation group.Finally,the clinical efficacy of each patient′s synchronized liver metastasis risk factor was calculated based on the nomogram.Results A total of 328 texture features were extracted from the T2WI.Seven most valuable features were selected after reducing the dimension by LASSO algorithm,including 3 co-occurrence matrices(GLCM)and 4 run-length matrices(RLM).Tumor staging and radiomic signatures were included in the Multifactor logistic regression to build the prediction model and nomogram.The accuracy of predicting SRLM was 0.862 and 0.844 in the training and the verification group,respectively.To evaluate the accuracy of the nomogram,radiomics signature and the tumor staging in all cases were 0.857,0.832 and 0.663,respectively.There was no significant difference in the number of SRLM cases between the high risk group and the low risk group based on nomogram(P>0.05).Conclusion The radiomics nomogram based on T2WI can be used as a quantitative tool to predict synchronous liver metast

关 键 词：直肠肿瘤 影像组学 同步肝转移 磁共振成像 列线图 

分 类 号：R735.37[医药卫生—肿瘤]