CT影像组学模型对食管癌术后个体化辅助治疗的评估价值  

The value of a CT-based radiomics model to individual adjuvant therapy for esophageal cancer after esophagectomy

作　　者：莫笑开 林少帆[2] 伍光恒 陈焯枝 郑旭峰[2] 胡泽桓 肖健宁 林黛英[2] 张水兴 MO Xiaokai;LIN Shaofan;WU Guangheng;CHEN Zhuozhi;ZHENG Xufeng;HU Zehuan;XIAO Jianning;LIN Daiying;ZHANG Shuixing(Department of Radiology, the First Affiliated Hospital of Jinan University, Guangzhou 510630, Guangdong, China;Department of Radiology, Shantou Central Hospital, Shantou 515041, Guangdong, China;College of Information Scientific and Technology, Jinan University, Guangzhou 510632, Guangdong, China)

机构地区：[1]暨南大学附属第一医院影像科,广东广州510630 [2]汕头市中心医院放射科,广东汕头515041 [3]暨南大学信息科学技术学院,广东广州510632

出　　处：《暨南大学学报（自然科学与医学版）》2022年第3期302-311,共10页Journal of Jinan University(Natural Science & Medicine Edition)

基　　金：国家自然科学基金项目(81871323);广东省自然科学基金项目(2018B030311024)。

摘　　要：目的:构建及验证基于影像组学预测食管鳞状细胞癌切除术后的无病生存期(DFS)模型,并进一步其评估术后辅助治疗的价值。方法:回顾性分析184例接受手术治疗的食管鳞状细胞癌患者。放射科医师手工勾画患者治疗前CT图像上的肿瘤感兴趣区后提取影像组学特征,采用最小绝对收缩和选择算子(LASSO)-Cox回归筛选关键特征并构建影像组学模型。同时,通过临床特征筛选及建立临床模型,结合以上模型后得到影像组学-临床模型。采用一致性指数(C指数)评价模型的预测效能。结果:影像组学模型由10个影像组学特征构成。病理淋巴结分期为DFS独立的临床风险因素。影像组学-临床模型表现出对肿瘤进展高低风险人群的良好区分度,训练集C指数为0.744,95%置信区间(CI)为0.689~0.799,验证集C指数为0.774,95%CI为0.676~0.872。高风险组中位DFS明显小于低风险组其中训练集(16个月vs 39个月,P<0.001);验证集(11个月vs 24个月,P<0.001)。同时,在训练集的低风险患者中,手术后辅助治疗患者与单纯手术患者的DFS比较无显著统计学差异(DFS分别为34.4个月,32.5个月,P=0.816)。结论:本研究构建并验证了基于治疗前CT图像预测食管鳞状细胞癌患者DFS的影像组学-临床模型,该模型具有评估术后肿瘤进展低风险人群的辅助治疗的价值。Objective:To develop and validate the radiomic based-model to predict disease-free survival(DFS)of esophageal squamous cell carcinoma after surgery and further evaluate the value of postoperative adjuvant therapy.Methods:A total of 184 patients with esophageal squamous cell carcinoma(ESCC)who underwent esophagectomy were analyzed retrospectively.Radiologists manually outlined the tumor region of interest on the pretreatment CT images.Then the radiomic features were extracted,and the least absolute shrinkage and selection operator(LASSO)-Cox regression was used to select the significant radiomic features and construct the radiomic model.Meanwhile,the clinical model was built based on the clinical data.A radiomic-clinical model was obtained by combining the radiomic and clinical models.The concordance index(C-index)was used to evaluate the predictive performance of the models.Results:The radiomic model consisted of 10 significant radiomic features.Pathological lymph node staging was an independent clinical risk factor for DFS.The radiomic-clinical model demonstrated good performance in stratifying the high-risk and low-risk groups,yielding the C index of 0.744(95%CI:0.689~0.799)and 0.774(95%CI:0.676~0.872)in the training and validation cohorts respectively.The high-risk and the low-risk groups of disease progression were well-identified in the training and validation cohorts.The median DFS was significantly shorter in the high-risk group than in the low-risk group(training cohort:16 months vs.39 months,P<0.001;validation cohort:11 months vs.24 months,P<0.001).Additionally,among the low-risk patients in the training cohort,there was no significant statistical difference in the DFS between patients with postoperative adjuvant therapy and surgery alone(DFS:34.4 months,vs.32.5 months,respectively,P=0.816).Conclusion:Our study has developed and validated a radiomic-clinical model for predicting DFS in patients with esophageal squamous cell carcinoma based on pretreatment CT images.The model can assess the value of adjuv

关 键 词：食管鳞状细胞癌 无病生存期 辅助治疗 计算机断层扫描 影像组学 

分 类 号：R445.3[医药卫生—影像医学与核医学]