Nimg-59. Radiologic Subtypes of Glioblastoma Calculated Via Multi-parametric Imaging Signatures Reveal Complementary Information to Current Who Classification
Year: 2017
Type: article
Study design: Observational Study
Abstract: The World Health Organization (WHO) has recently changed the classifications of glioblastoma based on molecular information. The aim of this study was to systematically find imaging signatures via quantitative multi-parametric MRI pattern analysis that can provide complementary information to the current WHO classifications for patients’ prognosis. We analyzed a discovery cohort of N=208 de novo glioblastoma patients with available preoperative multi-parametric MRI data (T1, T1-Gd, T2, FLAIR, DTI, DSC). We conducted a pattern analysis based on comprehensive features, which were extracted from multi-parametric MRI in various parts of the tumor. The features comprised shape measures of circularity and sphericity, Haralick texture measures, intensity statistical parameters, volumetric measures, and percentage in each bin of intensity distribution. Unsupervised clustering was used on imaging measures to find radiologic subtypes, which were initially cross-validated in the discovery cohort and subsequently evaluated in a replication cohort (N=53). Our results revealed three radiologic subtypes of glioblastoma with distinct overall survival rates (hazard ratio = 4.174 (2.463–7.072,P<0.001), CI 95%), spatial location, and radiological measures of cell-density, vascularization, infiltration, and size of peritumoral edema (ANOVA: P<0.01 for all). The combination of subtype with IDH1 mutation predicted better survival compared to either of these alone (Subtype=73.91%, IDH1=75.00%, Combined=80.02% classification accuracy). Further, our subtypes were not only consistent with WHO classifications by assigning 75% IDH1 mutants to the long survivor subtype, but were also an excellent predictor of survival within IDH1-wildtype patients (median survival: 7, 11.5, 19 months for IDH1-wildtype patients in three subtypes, ANOVA P<0.01), thereby highlighting the complementary value of imaging to the established WHO classifications for patient’s prognosis. Advanced pattern analysis of multi-parametric MRI reveals radiological subtypes of glioblastoma that are predictive of an individual patient’s prognosis, substantially beyond the current IDH1-based WHO classifications and which therefore might assist in personalized treatment.
Source: Neuro-Oncology
Institution: University of Pennsylvania
Language: English
Subfield: Radiology, Nuclear Medicine and Imaging
Field: Medicine
Domain: Health Sciences
Open Access status: green
OpenAlex ID: W2768026820