基于3D-MRI影像组学在成人脑弥漫性胶质瘤IDH分型的应用价值

Application Value of 3D-MRI Radiomics in IDH Classification of Adult Diffuse Glioma

目的:建立基于三维磁共振成像(3D-MRI)的影像组学预测成人弥漫性胶质瘤异柠檬酸脱氢酶(IDH)分子分型的方法,以辅助术前决策和预后评估。方法:回顾性收集在深圳大学第一附属医院2016年7月至2021年1月期间经手术切除病理证实的66例成人脑弥漫型胶质瘤患者的MRI图像,临床病理及IDH分子分型资料。IDH分子分型采用聚合酶链反应(qPCR)测序法。基于患者T1WI、T2WI图像勾画3D感兴趣区(ROI),提取肿瘤影像组学特征。将数据以7:3随机分为训练集和测试集,用Spearman相关性分析和LASSO对特征进行降维,筛选出与IDH分子分型相关的影像组学特征,并建立预测IDH分型模型,采用受试者工作特征曲线(ROC)、混淆矩阵验证预测模型性能。结果:本研究共纳入66例成人弥漫性胶质瘤,其中IDH突变型27例,IDH野生型39例,每例患者影像图像提取396个特征,筛选出6个相关的影像组学特征建立IDH分型预测模型,训练集的曲线下面积(AUC)为0.882,测试集的AUC为0.841;训练集的准确度、灵敏度、特异度分别为0.894、0.842、0.929,测试集的准确度、灵敏度、特异度分别为0.789、0.750、0.818。结论:基于MRI的影像组学能够预测成人脑弥漫性胶质瘤,可无创性获得IDH分型结果,对于术前/无法手术者制定诊治方案及随诊策略具有一定的临床价值。

Objective To establish a radiomics method based on three-dimensional magnetic resonance imaging(3D-MRI) to predict the molecular typing of isocitrate dehydrogenase(IDH) in adult diffuse gliomas, so as to assist preoperative decision-making and prognosis assessment. Methods MRI images, clinicopathological and IDH molecular typing data of 66 adult patients with diffuse glioma confirmed by surgical resection and pathology in the First Affiliated Hospital of Shenzhen University from July 2016to January 2021 were retrospectively collected. IDH molecular typing was performed by quantitative polymerase chain reaction(qPCR)sequencing. 3D regions of interest(ROI) were delineated based on T1WI、T2WI images of patients, and their imaging features were extracted. The data were randomly divided into training set and test set at 7:3. Spearman correlation analysis and LASSO were used to reduce the dimensionality of the features, and the radiomics features related to IDH molecular typing were screened out.The IDH predictive model was established and the performance of the prediction model was verified by ROC curve and confusion matrix. Results A total of 66 cases of adult diffuse glioma were included in this study, including 27 IDH mutant cases and 39 IDH wild type cases. 396 features were extracted from the images of each patient, and 6 related features were screened out to establish IDH prediction model. The area under curve(AUC) of the training set was 0.882. The AUC of the test set was 0.841. The accuracy,sensitivity and specificity of the training set were 0.894, 0.842 and 0.929, respectively. The accuracy, sensitivity and specificity of the test set were 0.789, 0.750 and 0.818, respectively. Conclusion MRI-based imaging can predict diffuse glioma in adult brain and obtain IDH classification results in a non-invasive way, which has certain clinical value for preoperative and inoperable patients to make diagnosis and treatment plans and follow-up strategies.