Bumes Elisabeth, Wirtz Fro-Philip, Fellner Claudia, Grosse Jirka, Hellwig Dirk, Oefner Peter J, Häckl Martina, Linker Ralf, Proescholdt Martin, Schmidt Nils Ole, Riemenschneider Markus J, Samol Claudia, Rosengarth Katharina, Wendl Christina, Hau Peter, Gronwald Wolfram, Hutterer Markus
Department of Neurology and Wilhelm Sander-NeuroOncology Unit, Regensburg University Hospital, 93053 Regensburg, Germany.
Institute of Functional Genomics, University of Regensburg, 93053 Regensburg, Germany.
Cancers (Basel). 2020 Nov 17;12(11):3406. doi: 10.3390/cancers12113406.
( mutation is an important prognostic factor and a potential therapeutic target in glioma. Immunohistological and molecular diagnosis of mutation status is invasive. To avoid tumor biopsy, dedicated spectroscopic techniques have been proposed to detect D-2-hydroxyglutarate (2-HG), the main metabolite of , directly in vivo. However, these methods are technically challenging and not broadly available. Therefore, we explored the use of machine learning for the non-invasive, inexpensive and fast diagnosis of status in standard H-magnetic resonance spectroscopy (H-MRS). To this end, 30 of 34 consecutive patients with known or suspected glioma WHO grade II-IV were subjected to metabolic positron emission tomography (PET) imaging with O-(2-F-fluoroethyl)-L-tyrosine (F-FET) for optimized voxel placement in H-MRS. Routine H-magnetic resonance (H-MR) spectra of tumor and contralateral healthy brain regions were acquired on a 3 Tesla magnetic resonance (3T-MR) scanner, prior to surgical tumor resection and molecular analysis of status. Since 2-HG spectral signals were too overlapped for reliable discrimination of mutated () and wild-type () glioma, we used a nested cross-validation approach, whereby we trained a linear support vector machine (SVM) on the complete spectral information of the H-MRS data to predict status. Using this approach, we predicted status with an accuracy of 88.2%, a sensitivity of 95.5% (95% CI, 77.2-99.9%) and a specificity of 75.0% (95% CI, 42.9-94.5%), respectively. The area under the curve (AUC) amounted to 0.83. Subsequent ex vivo H-nuclear magnetic resonance (H-NMR) measurements performed on metabolite extracts of resected tumor material (eight specimens) revealed myo-inositol (M-ins) and glycine (Gly) to be the major discriminators of status. We conclude that our approach allows a reliable, non-invasive, fast and cost-effective prediction of status in a standard clinical setting.
(突变是胶质瘤的一个重要预后因素和潜在治疗靶点。免疫组织学和分子诊断突变状态具有侵入性。为避免肿瘤活检,已提出专门的光谱技术直接在体内检测D-2-羟基戊二酸(2-HG),即 的主要代谢产物。然而,这些方法在技术上具有挑战性且未广泛应用。因此,我们探索了使用机器学习在标准氢磁共振波谱(H-MRS)中对 状态进行非侵入性、低成本且快速的诊断。为此,34例连续的已知或疑似WHO II-IV级胶质瘤患者中的30例接受了用O-(2-氟乙基)-L-酪氨酸(F-FET)进行的代谢正电子发射断层扫描(PET)成像,以优化H-MRS中的体素放置。在手术切除肿瘤和对 状态进行分子分析之前,在3特斯拉磁共振(3T-MR)扫描仪上获取肿瘤和对侧健康脑区的常规氢磁共振(H-MR)波谱。由于2-HG光谱信号重叠过多,无法可靠地区分 突变( )和 野生型( )胶质瘤,我们使用了嵌套交叉验证方法,即我们在H-MRS数据的完整光谱信息上训练线性支持向量机(SVM)以预测 状态。使用这种方法我们预测 状态的准确率为88.2%,灵敏度为95.5%(95%CI,77.2 - 99.9%),特异性为75.0%(95%CI,42.9 - 94.5%)。曲线下面积(AUC)为0.83。随后对切除的肿瘤材料(8个样本)的代谢物提取物进行的离体氢核磁共振(H-NMR)测量显示,肌醇(M-ins)和甘氨酸(Gly)是 状态的主要鉴别指标。我们得出结论,我们的方法能够在标准临床环境中对 状态进行可靠、非侵入性、快速且经济高效的预测。 )
需注意,原文中部分内容缺失关键信息(如 具体指代不明),翻译时保留了原文格式。