基于MRI影像基因组学预测肺腺癌脑转移瘤EGFR突变状态  

Radiogenomics Models Based on MRI of Brain Metastasis for Predicting EGFR Mutation Status in Lung Adenocarcinoma

作　　者：隋莲玉 王佳宁[1] 任嘉梁 蔡静薇 张宇[1] 殷小平[1] SUI Lianyu;WANG Jianing;REN Jialiang(Department of Radiology,the Affiliated Hospital of Hebei University,Baoding,Hebei Province 071000,P.R.China)

机构地区：[1]河北大学附属医院放射科,保定071000 [2]河北大学临床医学院,保定071000 [3]通用电气药业(上海)有限公司,上海2100003

出　　处：《临床放射学杂志》2023年第1期20-25,共6页Journal of Clinical Radiology

基　　金：河北大学研究生创新资助项目(编号:HBU2022ss024);河北大学优秀青年科研创新团队(编号:605020521007);河北大学医学学科培育项目(编号:2021B19)。

摘　　要：目的基于脑转移瘤的MRI影像组学特征预测原发肺腺癌中表皮生长因子受体(EGFR)突变状态,并探讨预测EGFR突变的最佳MRI序列。方法回顾性分析146例肺腺癌脑转移患者(EGFR突变型104例,野生型42例)的增强T_(1)WI序列、FLAIR序列和DWI序列图像,按7∶3的比例随机分为训练集(103例)和验证集(43例),基于以上3个MRI序列进行影像组学特征提取及预测模型构建,然后利用验证集数据评价组学模型的效能。结果在增强T_(1)WI、FLAIR及DWI联合序列的训练组中,从总共3111个组学特征中最终筛选出来的9个显著特征,联合多序列构建的逻辑回归预测模型训练集曲线下面积为0.830(95%CI 0.748~0.913),准确度为0.718,敏感度为0.644,特异度为0.900;在验证集中,曲线下面积为0.823(0.690~0.955),准确度为0.744,敏感度为0.710,特异度为0.833。结论基于MRI多个联合序列的影像组学特征可作为预测肺腺癌EGFR突变状态的无创辅助工具。Objective Identification of epidermal growth factor receptor(EGFR)mutation status in primary lung adenocarcinoma basing magnetic resonance imaging(MRI)multi-sequences radiogenomics features of brain metastasis and to explore the optimal single sequence and co-sequences for prediction.Methods Contrast-enhanced T1-weighted images(T_(1)WI),fluid-attenuated inversion recovery(FLAIR)and diffusion weighted imaging(DWI)sequences of 146 patients with brain metastasis from lung adenocarcinoma(104 with mutant EGFR,42 with wild-type EGFR)were selected for radiomics features extraction retrospectively.Brain metastasis dataset from all sequences were randomly divided into training set(n=103)and testing set(n=43)according to the ratio of 7∶3.Based on the above three MRI sequences,radiomics feature extraction and prediction models were performed in the training cohort,and then the radiomics signature performance was evaluated using the validation cohort.Results The radiomics signature was built with the Logistic regression model based on 9 selected features from 3111 radiomics features in the training cohorts for contrast-enhanced T_(1)WI,FLAIR and DWI.Logistic regression model constructed by combining multiple sequences yielded an AUC of 0.830(95%CI 0.748-0.913),a classification accuracy of 0.718,sensitivity of 0.644 and specificity of 0.900 in the training cohort.The AUC was 0.823(0.690-0.955),accuracy of 0.744,sensitivity of 0.710 and specificity of 0.833in the testing dataset.Conclusion We developed a MR multiple joint sequences radiomics signature that can be used as a noninvasive auxiliary tool for predicting EGFR mutation status in lung adenocarcinoma.

关 键 词：肺腺癌 脑转移瘤 影像基因组学 磁共振成像 表皮生长因子受体 

分 类 号：R734.2[医药卫生—肿瘤] R445.2[医药卫生—临床医学]