Multiparametric Evaluation in Differentiating Glioma Recurrence from Treatment-Induced Necrosis Using Simultaneous 18F-FDG-PET/MRI: A Single-Institution Retrospective Study

Abstract

Background and purpose: Differentiating glioma recurrence from treatment-induced necrosis can be a challenge on conventional imaging. This study aimed to assess the diagnostic performance of each functional MR imaging and PET parameter derived by using simultaneous FDG-PET/MR imaging individually and in combination in the evaluation of suspected glioma recurrence.

Materials and methods: Thirty-five treated glioma patients with 41 enhancing lesions (World Health Organization grade II = 9, III = 13, IV = 19) on MR imaging after an operation followed by radiation therapy and/or chemotherapy formed part of this study. Using PET/MR imaging, we calculated the normalized mean relative CBV, mean ADC, Cho/Cr, and maximum and mean target-to-background ratios. Statistical analysis was performed to determine the diagnostic performance of each parameter by receiver operating characteristic analysis individually and in combination with multivariate receiver operating characteristic analysis for the detection of glioma recurrence. Histopathology or clinicoradiologic follow-up was considered the criterion standard.

Results: Of 35 patients, 25 (30 lesions) were classified as having a recurrence and 10 (11 lesions) patients as having treatment-induced necrosis. Parameters like rCBV_mean (mean relative CBV), ADC_mean, Cho/Cr, and maximum and mean target-to-background ratios were statistically significant in the detection of recurrent lesions with an accuracy of 77.5%, 78.0%, 90.9%, 87.8%, and 87.8%, respectively. On multivariate receiver operating characteristic analysis, the combination of all 3 MR imaging parameters resulted in an area under the curve of 0.913 ± 0.053. Furthermore, an area under the curve of 0.935 ± 0.046 was obtained when MR imaging parameters (ADC_mean and Cho/Cr) were combined with the PET parameter (mean target-to-background ratio), demonstrating an increase in diagnostic accuracy.

Conclusions: Simultaneous PET/MR imaging with FDG offers correlative and synergistic multiparametric assessment of glioma recurrence with increased accuracy and clinical utility.

Fig 1.

PET/MR imaging of a 73-year-old man with posttreatment (operation, radiation therapy, and chemotherapy) right temporoparietal glioblastoma multiforme with suspected recurrence proved to be treatment-induced necrosis on 11-month PET/MR imaging follow-up. Axial T1-weighted postcontrast image (A) shows an enhancing lesion along the margins of the operated bed with a freehand ROI drawn defining the enhancing component, which was copied and pasted on the FDG image (B), PET/MR fused image (C), CBV map (D), and ADC map (E) to derive SUV_max and SUV_mean, CBV_mean, and ADC_mean show no focal increased FDG uptake, CBV, and diffusion restriction on the ADC map. Multivoxel ¹H-MR spectroscopy (F) along the enhancing margin shows no increased Cho/Cr ratio.

Fig 2.

PET/MR images of a 49-year-old woman with posttreatment (operation and radiaton therapy) left frontotemporal anaplastic oligodendroglioma that proved to be a recurrence on histopathologic examination (glioblastoma multiforme with an oligodendroglial component; World Health Organization grade IV with a large area of necrosis). Axial T1-weighted postcontrast image (A) shows an enhancing lesion in the tumor bed and involving the corpus callosum. Freehand ROI drawn defining the enhancing component of the lesion and copied and pasted on the FDG image (B), the PET/MR fused image (C), the CBV map (D), and the ADC map (E) to derive SUV_max and SUV_mean, CBV_mean, and ADC_mean shows increased FDG uptake, CBV, and diffusion restriction in the ADC map. Multivoxel ¹H-MR spectroscopy (F) obtained on the FDG avid enhancing area shows an increased Cho/Cr ratio. The enhancing region anterior to the target lesion has no FDG uptake and no increased CBV, and diffusion restriction represents necrosis (white arrow).

Fig 3.

Box-and-whisker plots comparing rCBV_mean (A), ADC_mean (B), Cho/Cr (C), and TBR_mean (D) between the glioma recurrence and treatment-induced necrosis. Whiskers represent the range of data; boxes represent the distance between the first and third quartiles.

Fig 4.

Receiver operating characteristic curves with their respective AUC values of MR imaging parameters (A) showing the high diagnostic performance of Cho/Cr in the detection of glioma recurrence. With multivariate ROC analysis, the ROC curve and AUC of all 3 MR imaging parameters combined show a significant increment in AUC over the individual MR imaging parameters (B). ROC curves with their respective AUC values of FDG parameters (C) show the high diagnostic performance of TBR_mean in the detection of glioma recurrence. Multivariate ROC analysis, ROC curve, and AUC of the best performing FDG-PET/MR imaging combination of ADC_mean, Cho/Cr, and TBR_mean show a significant increment over individual MR imaging or PET parameters (D).