Patient population
This retrospective study was approved by our Institutional Review Board, and the requirement of informed consent was waived. From November 2021 to July 2023, sixty-three consecutive patients with pathologically confirmed primary IDH-wildtype GBM were searched from the picture archiving and communication system in our hospital. The inclusion criteria for this study are shown below: (1) Available preoperative MRI images; (2) Confirmed histopathological diagnosis of IDH-wildtype GBM; (3) Age ≥ 18 years old. The excluded criteria for this study are shown below: (1) Any puncture or therapeutic intervention (n=9); (2) Image analysis is not possible due to image motion artifacts (n=3); (3) Absence of definitive information of the pMGMT methylation status (n=5). Finally, forty-six primary IDH-wildtype GBM patients with known pMGMT methylation status (25 unmethylated and 21 methylated) were enrolled in our study.
Data acquisition
A 3.0T superconducting MRI scanner (Siemens Verio, Erlangen, Germany) is used for image acquisition. The acquired images included T1-weighted imaging (WI), T2WI, and contrast-enhanced T1WI (CET1). The detailed scanning parameters are shown as follows: T1WI (repetition time [TR] 250 ms, echo time [TE] 2.48 ms, field of view [FOV] 220 mm×220 mm, slice thickness 5 mm, interslice gap 1.0 mm, and matrix 256 × 256.); T2WI ( TR 4000 ms, TE 96 ms, FOV 220 mm×220 mm, slice thickness 5 mm, interslice gap 1.0 mm, and matrix 256 × 256.); The CET1 was obtained with the same parameters as for T1WI after the contrast agent (Bayer Schering Pharma AG, Berlin, Germany) was administered at a dose of 0.1 mmol/kg at a flow rate of 3.0 ml/s via an intravenously.
Image analysis
Two experienced neuroradiologists (with 15 and 20 years, respectively) reviewed all images, blinded to pathological information. Conventional MRI signs, including location (frontal, temporal, parietal, posterior fossa, or others), across the midline (yes or no), margin (clear or fuzzy), necrosis/cystic changes (yes or no), hemorrhage (yes or no), and enhancement pattern (ring enhancement or non-ring enhancement) were assessed and recorded. An open-source FireVoxel software (New York University, NY, USA; https://www.firevoxel.org/) was applied to perform histogram analysis. Firstly, a region of interest (ROI) including all components of the lesion was drawn on the axial CET1 images. Subsequently, FireVoxel software automatically generates and extracts histogram parameters (including minimum, maximum, mean, 1st (Perc.01), 5th (Perc.05), 10th (Perc.10), 25th (Perc.25), 50th (Perc.50), 75th (Perc.75), 90th (Perc.90), 95th (Perc.95), and 99th (Perc.99) percentiles as well as standard deviation (SD), variance. coefficient of variation (CV), skewness, kurtosis, and entropy).
Histopathological analysis
The histopathological evaluation of the resected surgical specimens was conducted by a neuropathologist with 15 years of experience. Histopathological diagnosis was performed by the newest WHO 2021 Classification of Tumors of the Central Nervous System (CNS). The pMGMT methylation status was estimated with a methylation-specific polymerase chain reaction, which is described in the literature [13].
Statistical analysis
Statistical analyses were performed with SPSS (SPSS Version 27.0, Chicago, IL, USA). A p-value less than 0.05 was considered statistically significant. The categorical variables were compared using the χ2 test, while noncategorical variables were analyzed using an independent t-test (normal distribution) or the Man-Whitney U test (non-normal distribution), respectively. The diagnostic performance of CET1 histogram parameters was assessed by receiver operating characteristic (ROC) analysis. Delong’s test was used to compare the difference between areas under the curve (AUCs). In addition, inter-observer reliability of histogram parameters was measured by the intraclass correlation coefficient (ICC) test .