影像组学对进展期远端胃癌下切缘预测的应用研究  

Application study of imagingomics in prediction of distal margin of advanced distal gastric cancer

作　　者：张超[1] 张训营 刘磊[1] 亓伟 王东升[1] 卢云[1,2] 周晓明 王涛[1] 牛田野[3,4] 张宪祥[1] ZHANG Chao;ZHANG Xunying;LIU Lei;QI Wei;WANG Dongsheng;LU Yun;ZHOU Xiaoming;WANG Tao;NIU Tianye;ZHANG Xianxiang(Affiliated Hospital of Qingdao University,Shandong Qingdao 266000,China;Shandong Provincial Key Laboratory of Digital Medicine and Computer Assisted Surgery,Shandong Qingdao 266000,China;Institute of Translational Medicine,Zhejiang University,Zhejiang Hangzhou 310020,China;Department of Radiology,Zhejiang University Medical College Affiliated Sir Run Shaw Hospital,Zhejiang Hangzhou 310016,China)

机构地区：[1]青岛大学附属医院,山东青岛266000 [2]山东省数字医学与计算机辅助手术重点实验室,山东青岛266000 [3]浙江大学转化医学研究所,浙江杭州310020 [4]浙江大学医学院附属邵逸夫医院放射科,浙江杭州310016

出　　处：《现代肿瘤医学》2022年第17期3203-3208,共6页Journal of Modern Oncology

基　　金：国家自然科学基金青年基金(编号:81802473);青岛大学附属医院青年科研基金项目(编号:3458)。

摘　　要：目的:评估术前增强CT影像组学特征对进展期远端胃癌术中下切缘阳性的预测价值。方法:回顾性分析183例进展期远端胃癌患者术前增强CT图像,在CT图像上沿病变边缘手动绘制感兴趣区域(ROI),并提取影像组学特征。利用Pearson相关性分析和序列前向浮动选择(SFFS)算法筛选特征,并构建影像组学模型、验证模型可靠性。此外,还开发了临床病理学模型,运用受试者工作特征曲线(ROC曲线)及曲线下面积(AUC值)对两个模型诊断性能进行比较。结果:筛选出5种最优影像组学特征,构建的影像组学模型表现出良好的诊断性能(AUC值=0.79),且比临床病理学模型稍好(临床病理学模型AUC值=0.61)。结论:影像组学模型在术前无创预测术中切缘状态方面表现出良好的潜力。Objective:To evaluate the predictive value of preoperative contrast-enhanced CT imaging in patients with advanced distal gastric cancer.Methods:The preoperative enhanced CT images of 183 patients with advanced distal gastric cancer were retrospectively analyzed.The region of interest(ROI)was manually drawn along the lesion edge on the CT images and the imaging features were extracted.The features were screened by Pearson correlation analysis and sequence forward floating selection(SFFS)algorithm,and the image omics model was constructed to verify the reliability of the model.In addition,a clinicopathological model was developed,and the diagnostic performance of the two models was compared using receiver operating characteristic(ROC)curve and area under the curve(AUC value).Results:Five optimal radiomic features were screened out,and the constructed radiomic model showed good diagnostic performance(AUC=0.79),which was slightly better than the clinicopathological model(AUC=0.61).Conclusion:The imaging omics model has good potential for noninvasive prediction of intraoperative margin status.

关 键 词：进展期远端胃癌 影像组学 术中切缘 计算机断层扫描(CT) 

分 类 号：R735.2[医药卫生—肿瘤]