首页 | 本学科首页   官方微博 | 高级检索  
检索        

基于CT影像组学术前预测胃癌淋巴血管侵犯
引用本文:李武超,陈琦,蒋仪,王荣品.基于CT影像组学术前预测胃癌淋巴血管侵犯[J].中国医学影像技术,2019,35(7):1057-1060.
作者姓名:李武超  陈琦  蒋仪  王荣品
作者单位:贵州省人民医院放射科, 贵州 贵阳 550002;贵州省智能医学影像分析与精准诊断重点实验室, 贵州 贵阳 550002,贵州省人民医院信息科, 贵州 贵阳 550002,贵州省人民医院信息科, 贵州 贵阳 550002,贵州省人民医院放射科, 贵州 贵阳 550002;贵州省智能医学影像分析与精准诊断重点实验室, 贵州 贵阳 550002
基金项目:贵州省科技计划项目(黔科合基础[2016]1096)、贵州省高层次创新型人才培养(GZSYQCC[2015]001)、贵州省人民医院博士基金(GZSYBS[2015]02)、贵州省人民医院青年基金(QZSYQN[2015]01)。
摘    要:目的 探讨基于CT影像组学术前预测胃癌淋巴血管侵犯的价值。方法 回顾性收集经手术病理证实的181例胃癌患者,将其随机分为训练集(n=120)和验证集(n=61)。首先基于增强CT静脉期图像分割肿瘤区域并提取影像组学特征;然后利用训练集筛选与淋巴血管侵犯相关特征,构建影像组学标签;最后基于验证集验证模型,采用ROC曲线及校准曲线评估模型的预测效能及拟合度。结果 最终提取7个与胃癌淋巴管血管侵犯最相关的影像组学特征构建影像组学标签,其在训练集的ROC曲线AUC为0.742P=0.001,95%CI(0.652,0.831)],验证集AUC为0.727P=0.002,95%CI(0.593,0.853)]。基于训练集所得最优阈值为0.422,模型在训练集中的准确率、敏感度和特异度分别为0.708、0.586、0.806,将此阈值用于验证集,其准确率、敏感度和特异度为0.689、0.519、0.824。校准曲线显示影像组学标签在训练集及验证集均具有较好的拟合度(P均>0.05)。结论 CT影像组学可作为预测胃癌术前淋巴血管侵犯提供的全新的无创影像学方法。

关 键 词:胃肿瘤  淋巴血管侵犯  体层摄影术  X线计算机  影像组学
收稿时间:2019/1/2 0:00:00
修稿时间:2019/5/8 0:00:00

CT radiomics for preoperatively predicting lymphovascular invasion of gastric cancer
LI Wuchao,CHEN Qi,JIANG Yi and WANG Rongpin.CT radiomics for preoperatively predicting lymphovascular invasion of gastric cancer[J].Chinese Journal of Medical Imaging Technology,2019,35(7):1057-1060.
Authors:LI Wuchao  CHEN Qi  JIANG Yi and WANG Rongpin
Institution:Department of Radiology, Guizhou Provincial People''s Hospital, Guiyang 550002, China;Key Laboratory of Intelligent Medical Image Analysis and Precision Diagnosis in Guizhou Province, Guiyang 550002, China,Department of Information, Guizhou Provincial People''s Hospital, Guiyang 550002, China,Department of Information, Guizhou Provincial People''s Hospital, Guiyang 550002, China and Department of Radiology, Guizhou Provincial People''s Hospital, Guiyang 550002, China;Key Laboratory of Intelligent Medical Image Analysis and Precision Diagnosis in Guizhou Province, Guiyang 550002, China
Abstract:Objective To investigate the value of CT radiomics for preoperative prediction of gastric cancer lymphovascular invasion. Methods Totally 181 patients with gastric cancer confirmed by surgical pathology were retrospectively collected and randomly divided into training set (n=120) and verification set (n=61). Firstly, the tumor area was delineated and segmented, and the radiomics features were extracted based on enhanced CT venous phase images. Then, the training set was used to screen features associated with lymphovascular invasion, and a radiomics signature was built. Finally, the model was validated based on the verification set, and ROC curve and calibration curve were used to assess the model''s predictive power and fit assessment. Results Seven radiomics features most relevant to lymphovascular invasion of gastric cancer were extracted and used to build the radiomics signature. The AUC of the training set was 0.742 (P=0.001, 95%CI0.652, 0.831]), of the verification set was 0.727 (P=0.002, 95%CI0.593, 0.853]). The optimal threshold based on the training set was 0.422. The accuracy, sensitivity and specificity of the model in the training set was 0.708, 0.586 and 0.806, respectively. This threshold was used for the verification set with accuracy, sensitivity, and specificity of 0.689, 0.519 and 0.824, respectively. The calibration curve showed that the radiomics signature had a good fit in both the training set and the verification set (both P>0.05). Conclusion CT radiomics can be used as a novel non-invasive imaging method for preoperatively predicting lymphovascular invasion in gastric cancer.
Keywords:stomach neoplasms  lymphovascular invasion  tomography  X-ray computed  radiomics
点击此处可从《中国医学影像技术》浏览原始摘要信息
点击此处可从《中国医学影像技术》下载免费的PDF全文
设为首页 | 免责声明 | 关于勤云 | 加入收藏

Copyright©北京勤云科技发展有限公司  京ICP备09084417号