摘要
目的探讨基于术前T2-液体衰减反转恢复序列(T2-FLAIR)图像建立的影像组学模型预测WHOⅡ~Ⅲ级胶质瘤Ki-67表达水平的价值。资料与方法回顾性分析2017年5月—2021年1月西南医科大学附属医院经术后病理证实的WHOⅡ~Ⅲ级胶质瘤114例,根据病理结果分为Ki-67高表达组63例和Ki-67低表达组51例,以7∶3随机分为训练组79例和验证组35例。使用3DSlicer软件在T2-FLAIR轴位图像上对病灶所有层面逐层手动勾画三维感兴趣区,包含瘤体及瘤周水肿。使用3DSlicer软件内Radiomics模块提取影像组学特征107个,分析筛除特征间相关系数>0.9的冗余特征,进一步使用最小绝对收缩和选择算法筛选特征并建立Logistic回归模型。采用受试者工作特征曲线对模型进行效能评价。结果最终筛选出6个影像组学特征,影像组学模型在训练组曲线下面积为0.916(95%CI 0.851~0.982),敏感度为91.4%,特异度为84.1%,准确度为87.3%,阳性预测值为82.0%,阴性预测值为92.5%;在验证组中曲线下面积为0.868(95%CI 0.735~1.000),敏感度为87.5%,特异度为84.2%,准确度为85.7%,阳性预测值为82.3%,阴性预测值为88.9%。结论基于术前T2-FLAIR图像建立的影像组学模型可有效预测WHOⅡ~Ⅲ级胶质瘤Ki-67表达水平。
Purpose To explore the value of a radiomics model based on preoperative T2-fluid attenuated inversion recovery(FLAIR)images in predicting Ki-67 index expression level in patients with WHO gradeⅡ-Ⅲgliomas.Materials and Methods A total of 114 patients with WHO gradeⅡ-Ⅲgliomas confirmed by pathology from May 2017 to January 2021 in the Affiliated Hospital of Southwest Medical University were analyzed retrospectively.According to the pathological results,all patients were divided into Ki-67 high expression group(n=63)and Ki-67 low expression group(n=51),and randomly divided into a training dataset(n=79)and a validation dataset(n=35)at the ratio of 7∶3.3D region-of-interest was manually delineated slice-by-slice on the axial plane for T2-FLAIR images covering the tumor and peripheral edema area via the 3DSlicer software.A total of 107 texture features were extracted via Radiomics module in 3DSlicer software,and redundant features with correlation coefficients between features>0.9 were screened.A Logistic regression prediction model was finally constructed via least absolute shrinkage and selection operator.Receiver operating characteristic curve analysis was performed to evaluate the performance of the model.Results Six radiomics features were finally selected.The radiomics model showed good performance in predicting the Ki-67 index expression level in both the training dataset(area under curve 0.916,95%CI 0.851-0.982,sensitivity 91.4%,specificity 84.1%,accuracy 87.3%,positive predictive value 82.0%,negative predictive value 92.5%,respectively)and the validation dataset(area under curve 0.868,95%CI 0.735-1.000,sensitivity 87.5%,specificity 84.2%,accuracy 85.7%,positive predictive value 82.3%,negative predictive value 88.9%,respectively).Conclusion The radiomics model based on the preoperative T2-FLAIR images can effectively predict the Ki-67 expression level in patients with WHO gradeⅡ-Ⅲgliomas.
作者
樊建坤
程勇
黄欢
王菲菲
罗世刚
唐光才
FAN Jiankun;CHENG Yong;HUANG Huan;WANG Feifei;LUO Shigang;TANG Guangcai(Department of Radiology,the Affiliated Hospital of Southwest Medical University,Luzhou 646000,China;不详)
出处
《中国医学影像学杂志》
CSCD
北大核心
2023年第4期315-320,共6页
Chinese Journal of Medical Imaging
作者简介
通信作者:唐光才,168345315@qq.com。
