基于CT特征建立放射组学模型对非小细胞肺癌患者治疗效果的鉴别价值  

The identification value of radiomics model established based on CT features on the therapeutic effect of patients with NSCLC

作　　者：张亚超[1] 乔辉[2] 李泽 李鹏[2] ZHANG Yachao;QIAO Hui;LI Ze;LI Peng(Department of Radiology,Tianjin Third Censrul Hospital,Tianjin 300250,China;Imaging Center,Tianjin People's Hospital,Tianjin 300121,China;Department of Medical Imaging,Tianjin Huanhu Hospital,Tianjin 300350,China)

机构地区：[1]天津市第三中心医院分院放射科,天津300250 [2]天津市人民医院影像中心,天津300121 [3]天津市环湖医院医学影像科,天津300350

出　　处：《实用放射学杂志》2020年第7期1129-1132,共4页Journal of Practical Radiology

摘　　要：目的 基于CT特征建立放射组学模型对非小细胞肺癌(NSCLC)患者治疗效果的鉴别价值.方法 回顾性分析进行立体定向体部放射治疗(SBRT)并符合纳入及排除标准的86例NSCLC患者资料.所有患者均进行随访,每位患者随访时间至少3个月,末次随访日期为2019-03-01.收集患者一般资料并在放疗前后进行CT检查,同时结合患者癌胚抗原(CEA)水平及肿瘤最大径建立多变量影像组学预测模型.通过Logistic回归分析筛选价值高的影像组学特征,通过绘制ROC曲线鉴别其对NSCLC患者治疗效果的预测价值.结果 死亡组患者肿瘤最大径、CEA异常率明显高于存活组(P<0.05);通过Lasso-logistic回归模型选取出18个系数非零特征;训练组敏感度为82.6%,特异度为75.6%,准确度为80.2%,AUC为0.787,验证组敏感度为64.3%,特异度为100%,准确度为86.4%,AUC为0.821;CEA水平、肿瘤最大径及影像组学标签是鉴别术后疗效的独立危险因素.结论 基于CT特征建立放射组学预测模型对NSCLC患者治疗效果的预测准确性较高,可作为预测及评估放疗疗效的客观指标.Objective To estabish a radiomies model besed on CT festures to identiy the therapeuie fee of nom small cell lung caneer(NSCLC)patients.Mlethuds A retrospective analyze of 86 patient:with NSCLC who underwent se reotactic body radintion therapy(SBRT)and met the inclusion and exclusion criteria was x rformed.All patiemts were ollowed u for at lenst 3 monhs,and the last fllow up date was March 1,2019.The general data of the patienrs were cleed and CT euminations were performed brefore and after radiotherapy.Then,combined with the level of carcinoembryonie antigen(CEA)and the largest diameter of the tumor,a.mulirariate imnaging omics prediction model was etablisood.The Logisie ragression smalysis was used t0 sereen the high walue imaging omies chrterisries,and it's predictive value to the treatment efeet on NSCLC patients wa idenified by poturig the ROC.Results The maximum diameter of the tumor and CEA abnormal rale in the death group were significantly higber than those in the survival group(P<0.05).18 coefficient:s nom xero festures were seleeted by Les80 logisier regression model.The sensitivity of the training group was 82.6%,the speifiaty was 75.6%,the aceuney was 80.2%。and the AUC was 0.787。while the snitivity of the verification group was 64.3%,the spcificity was 100%,the acuraey was 86.4%6。and the AUC was 0.821.The CEA leel,tumor ma ximal dianeter,and imaging omics labeling were independent risk faclors for identifying postoperative outcomes.Conclusion Extablisbhing radiomics predictin model based on CT feaures has high accuracy in predieting tbe therapeutic ffe of NSCI.C patients,and it can be used as an oejective index for predicting and evalusting the efficacy of radiotherapy.

关 键 词：计算机体层成像 放射组学模型 非小细胞肺癌 治疗效果 

分 类 号：R814.42[医药卫生—影像医学与核医学] R734.2[医药卫生—放射医学]