In Vivo Molecular Profiling of Human Glioma : Cross-Sectional Observational Study Using Dynamic Susceptibility Contrast Magnetic Resonance Perfusion Imaging.

To assess the diagnostic performance of dynamic susceptibility contrast perfusion magnetic resonance perfusion imaging (DSC-MRI) for in vivo human glioma molecular profiling. In this study 100 patients with histopathologically confirmed glioma who provided written informed consent were retrospectively assessed between January 2016 and February 2017 in two prospective trials that were approved by the local institutional review board. Cerebral blood volume (CBV) measurements from DSC-MRI were assessed, and histogram parameters of relative CBV (rCBV) results were compared among World Health Organization (WHO) 2016 based histological findings and molecular characteristics. A classification and regression tree (CART) algorithm with 10-fold cross-validation was used to calculate the diagnostic accuracy. The 90th percentile (C90) of rCBV was significantly lower in patients with the isocitrate dehydrogenase 1/2 (IDH1/2) mutation (2.86 ± 1.21; p < 0.001) and loss of alpha-thalassemia mental retardation syndrome X‑linked (ATRX) expression (2.23 ± 0.91; p < 0.001) than in those with the IDH1/2 wild type (4.78 ± 2.34) and maintained ATRX expression (4.30 ± 2.02). The standard deviation (SD) of rCBV was significantly higher in glioblastoma (GBM) with methylated O6-methylguanine DNA methyltransferase (MGMT; 1.99 ± 0.73; p = 0.001) than in those with unmethylated MGMT (1.20 ± 0.45). In CART analysis, rCBV predicted the molecular subgroup in 76.3% of astroglial tumors; however, the diagnostic performance was reduced to 48.1% by including oligodendrogliomas with chromosome 1p/19q co-deletion in the analysis due to substantial overlap of rCBV values between OD The DSC-MRI procedure may provide insight into the IDH1/2 mutation and ATRX expression status and MGMT methylation profile of diffuse glioma; however, taking integrated oligodendroglioma into account limits the diagnostic performance of rCBV in non-invasively predicting the molecular subtype.