The authors have declared that no competing interests exist.
Conceived and designed the experiments: DP MZ AR. Performed the experiments: DP MZ JEM JW AR. Analyzed the data: DP MZ JEM JW BW PK AR. Contributed reagents/materials/analysis tools: DP MZ JEM JW AR. Contributed to the writing of the manuscript: DP MZ JEM JW BW PB JON PK HPS WW AMN SH MEL MB AR.
Nuclear Overhauser Enhancement (NOE) mediated chemical exchange saturation transfer (CEST) is a novel magnetic resonance imaging (MRI) technique on the basis of saturation transfer between exchanging protons of tissue proteins and bulk water. The purpose of this study was to evaluate and compare the information provided by three dimensional NOE mediated CEST at 7 Tesla (7T) and standard MRI in glioblastoma patients.
Twelve patients with newly diagnosed histologically proven glioblastoma were enrolled in this prospective ethics committee–approved study. NOE mediated CEST contrast was acquired with a modified three-dimensional gradient-echo sequence and asymmetry analysis was conducted at 3.3ppm (B1 = 0.7 µT) to calculate the magnetization transfer ratio asymmetry (MTRasym). Contrast enhanced T1 (CE-T1) and T2-weighted images were acquired at 3T and used for data co-registration and comparison.
Mean NOE mediated CEST signal based on MTRasym values over all patients was significantly increased (p<0.001) in CE-T1 tumor (−1.99±1.22%), tumor necrosis (−1.36±1.30%) and peritumoral CEST hyperintensities (PTCH) within T2 edema margins (−3.56±1.24%) compared to contralateral normal appearing white matter (−8.38±1.19%). In CE-T1 tumor (p = 0.015) and tumor necrosis (p<0.001) mean MTRasym values were significantly higher than in PTCH. Extent of the surrounding tumor hyperintensity was smaller in eight out of 12 patients on CEST than on T2-weighted images, while four displayed at equal size. In all patients, isolated high intensity regions (0.40±2.21%) displayed on CEST within the CE-T1 tumor that were not discernible on CE-T1 or T2-weighted images.
NOE mediated CEST Imaging at 7T provides additional information on the structure of peritumoral hyperintensities in glioblastoma and displays isolated high intensity regions within the CE-T1 tumor that cannot be acquired on CE-T1 or T2-weighted images. Further research is needed to determine the origin of NOE mediated CEST and possible clinical applications such as therapy assessment or biopsy planning.
Magnetic resonance imaging (MRI) has become the gold standard for the assessment of intracerebral lesions and is thus the primary tool for diagnosis and follow up examination of glioblastoma
Chemical Exchange Saturation Transfer (CEST) imaging is a non-invasive MRI technique sensitive to endogenous mobile proteins and peptides respectively and their tissue specific concentration
Biomedical applications have for example been demonstrated for the detection and grading of tumors
At low saturation power (e.g. 0.6–0.8 µT) CEST studies revealed that saturation transfer at −2 to −5 ppm is predominantly mediated by Nuclear Overhauser Enhancement (NOE) effects
In the current study, we investigated if NOE-weighted CEST-MRI with high 3D spatial resolution at 7T and precise sequence co-registration provides additional information about glioblastoma imaging, specifically the visualization of surrounding tumor hyperintensities and isolated CEST high intensity regions (HIR) that do not display on CE-T1 or T2-weighted images.
Twelve patients (3 female, 9 male; age: 62.58±12.67 years) with newly diagnosed and subsequently histopathologically confirmed glioblastoma were included in this prospective study. The study was approved by the Medical Ethics Committee (Faculty of Clinical Medicine, University of Heidelberg, Germany) and written informed consent was received from all participants before enrollment.
CE-T1 weighted (TE = 4.04 ms, TR = 1710 ms, FoV 256×256, resolution 512×512, slice thickness 1 mm) and T2-weighted (TE = 89 ms, TR = 5140 ms, FoV 172×229, resolution 384×230, slice thickness 4 mm) images were acquired on a 3T whole body MR imaging system (Magnetom Verio/Trio TIM; Siemens Healthcare, Erlangen, Germany).
The CEST sequence was performed on a 7T whole body MRI scanner (Magnetom 7T; Siemens Healthcare, Erlangen, Germany) with a time delay of 1–5 days in relation to 3T MRI. CEST imaging was performed with a centric-reordered three-dimensional gradient echo sequence
Analyses of 3D-coregistered CE-T1, T2-weighted and CEST images were performed by two neuroradiologists (AR and PK). Discrepancies were resolved by consensus reading. CE-T1 tumor and tumor necrosis were identified on CE-T1 and peritumoral edema on T2-weighted images. Size and structure of the edema on T2-weighted images were visually compared to corresponding peritumoral hyperintensities on co-registered CEST images in three dimensions. The appearance of isolated CEST HIR on MTRasym within the area of CE-T1 tumor and within the area of peritumoral edema according to its extent on T2-weighted images were evaluated. Illustration of CEST HIR was performed in the same window (−10% to +5%) but with different color gradients for improved visualization. Finally, the appearance of tumor satellite lesions (defined as contrast enhanced lesions on CE-T1, diameter <1 cm, without connection to the main tumor) was investigated on CE-T1, T2-weighted and on corresponding CEST images.
Six regions of interest (ROI) were selected for each patient on co-registered data in a representative slice for quantitative MTRasym signal analysis. The selection of the ROIs was performed based on best visibility of the several tissues on the following sequences: 1) CE-T1 tumor and 2) tumor necrosis were selected on CE-T1 images. 3) Isolated CEST HIR within the CE-T1 tumor and 4) peritumoral CEST hyperintensites (PTCH) within T2 edema margins were directly selected on MTRasym. 5) Cerebrospinal fluid (CSF) and 6) contralateral normal appearing white matter (CLNAWM) were selected on T2-weighted images. For each ROI the average MTRasym was determined. Furthermore, the contribution of up and downfield effects on MTRasym within Z-spectra were visually compared for each ROI.
The data from ROI analysis was used for statistical evaluation. A repeated measures analysis of variance (rm ANOVA) for all regions and patients and post hoc Holm-Sidac pairwise multiple comparisons were performed with SigmaPlot version 12.5 (Systat Software, Inc., San Jose California USA). The level of significance was set at P<0.05.
CEST effects given by MTRasym were observed in a minimum-to-maximum range from −25% to +12% resulting from the asymmetry analysis based on the measured Z-spectra. 98.51% of all intracranial values were in the range from −12% to +5%. All tumors could be identified on MTRasym as hyperintense lesions since NOE mediated CEST effects decreased in all glioblastoma tumors. Highest MTRasym intensity values appear in CSF and in isolated CEST HIR of the CE-T1 tumor, both showing MTRAsym values of approximately 0. For CSF this is because no saturation transfer is apparent neither at +3.3 ppm nor at −3.3 ppm. Within the isolated CEST HIR of the CE-T1 tumor, MTRasym = 0 reflects that NOE signals (−3.3 ppm) are of equal size as saturation transfer effects at the opposite side of the Z-spectrum (+3.3 ppm).
In eight out of 12 patients, the peritumoral hyperintensity on CEST was smaller than on T2-weighted images. In four patients, CEST displayed congruent areas. In two of the eight patients with smaller peritumoral hyperintensity, the CEST hyperintensity moderately exceeded the T2 edema in one direction. In comparison to the edema on T2-weighted sequences, peritumoral hyperintensities on CEST displayed an irregular border and subareas of different signal intensity. Furthermore, stria like structures could be identified on CEST images within peritumoral hyperintensities (
Left frontal glioblastoma in a 59 year old man at 3 Tesla, CE-T1 (A) and T2-weighted images (B). On the selected slice the CEST contrast at 7 Tesla, based on MTRasym (C), displays peritumoral hyperintensities at equal extent compared to the edema on T2-weighted images. In contrast to T2-weighted images, the CEST peritumoral hyperintensity displays an irregular border and subareas of different signal intensity.
Tumor necrosis on CEST appeared predominantly hyperintense compared to average signal in peritumoral hyperintensities. Within the corresponding area of CE-T1 tumor, CEST images revealed heterogeneous signal intensities. The MTRasym signal intensity in the CE-T1 tumor varied from isointense to peritumoral hyperintensities to isolated CEST HIR.
Isolated CEST HIR on MTRasym within the CE-T1 tumor could be observed in all 12 patients. In eight patients isolated CEST HIR could be additionally identified within the edema according to its extent on T2-weighted images.
A total of eight tumor satellites were identified in the patient collective on CE-T1. Four of these eight satellites were clearly hyperintense both on T2-weighted images and on CEST, while three of the eight CE-T1 satellites barely displayed on T2-weighted images and were also clearly visible on CEST images (
Tumor satellite of a glioblastoma subcortical temporal right in a 67 year old woman. The satellite presents a clear enhancement on CE-T1 (arrow in A) and barely displays on the T2-weighted image (arrow in B). In contrast, the satellite displays clearly hyperintense on CEST based on MTRasym (C) and matches with the area of contrast enhancement on the CE-T1 image (A). Furthermore also CSF in lateral ventricles and cerebral sulci displays hyperintense on MTRasym.
Patient No. | Size of peritumoral hyperintensity: | Appearance of isolated high intensity regions (HIR) on MTRasym in the area of: | Satellite lesions: | |
CEST versus T2-w. | CE–T1 tumor | T2 peritumoral edema | CEST hyperintense/Total on CE-T1 | |
#1 | smaller | Y | Y | ∅ |
#2 | equal | Y | N | 1/1 |
#3 | smaller | Y | N | ∅ |
#4 | equal | Y | Y | 2/2 |
#5 | smaller | Y | Y | 1/1 |
#6 | smaller | Y | Y | ∅ |
#7 | smaller | Y | N | ∅ |
#8 | smaller | Y | Y | ∅ |
#9 | smaller* | Y | N | ∅ |
#10 | equal | Y | Y | ∅ |
#11 | smaller* | Y | Y | 2/3 |
#12 | equal | Y | Y | 1/1 |
Peritumoral hyperintensity: Comparison of the extent of the peritumoral hyperintensity on CEST and T2-weighted images (smaller* = total extent smaller on CEST contrast but exceeding the margins of the T2 edema in one direction).
Mean MTRasym in CE-T1 tumor was −1.99±1.22% and −1.36±1.30% in tumor necrosis. For isolated CEST HIR on MTRasym within CE-T1 tumor the mean signal strength was 0.40±2.21% and −3.56±1.24% in PTCH within T2 edema margins. In CSF average MTRasym value was 0.76±1.29% and −8.38±1.19% in CLNAWM (
Left occipital glioblastoma of a 79 year old patient, CE-T1 (A) and T2-weighted images (B) with color coded ROIs: CE-T1 tumor, isolated CEST HIR within CE-T1 margins, tumor necrosis, PTCH within T2 edema margins, CSF and CLNAWM. CEST contrast based on MTRasym (C): Same ROIs illustrated in green for improved visualization. Z-spectrum (D) and asymmetry analysis (E) shown. Analyses of Z-spectra reveals that a decrease of NOE upfield effects at −3.3 ppm causes the hyperintense MTRasym contrast in the tumor regions, while no clear APT peak around +3.3 ppm could be identified in any of the analyzed tissues. Even though MTRasym shows high intensities both in CSF and isolated CEST HIR within CE-T1 tumor, Z-spectrum analysis reveals that the underlying asymmetry has a different origin: no saturation transfer is apparent in CSF at ±3.3 ppm (D black line) while in tumor regions (D dark green, dark blue and light blue lines) MTRasym = 0 reflects that NOE signals (−3.3 ppm) and saturation transfer effects at the opposite side of the Z-spectrum (+3.3 ppm) are of equal size. Furthermore the width of the Z-spectrum of CSF is decreased due to the longer T2 relaxation time.
The analysis of variance (ANOVA) with repeated measures was p<0.001 for statistically significant differences among the six groups by ROI analysis. Post hoc Holm-Sidac pairwise multiple comparisons showed that average MTRasym of CE-T1 tumor, isolated CEST HIR within the CE-T1 tumor, tumor necrosis, PTCH within T2 edema margins and CSF were all significantly higher than MTRasym of CLNAWM (p<0.001). Mean MTRasym in PTCH within T2 edema margins was significantly increased (p<0.001) compared to CLNAWM and significantly lower than in CE-T1 tumor (p = 0.015) and tumor necrosis (p<0.001). Average MTRasym in isolated CEST HIR within CE-T1 tumor was significantly higher than in the whole CE-T1 tumor (p<0.001) and PTCH within T2 edema margins (p<0.001). In tumor necrosis, MTRasym was significantly lower than in isolated CEST HIR within CE-T1 tumor (p = 0.007) and CSF (p = 0.001) but not significantly different compared to CE-T1 tumor (p = 0.42) (
Boxplots of mean MTRAsym values on CEST contrast over all patients (N = 12). Overall mean MTRasym (red stars) and outliers (red crosses) are additionally illustrated. MTRasym values in all tumor areas (CE-T1 tumor, isolated CEST HIR in CE-T1 tumor, tumor necrosis) and CSF are significantly higher than in CLNAWM (p<0.001). Average signal intensity in PTCH within T2 edema margins is significantly higher (p<0.001) than in CLNAWM and significantly lower (p = 0.015) than in CE-T1 tumor and tumor necrosis (p<0.001). The whiskers of the boxplot for isolated CEST HIR indicate a high variance within this group, which is due to smaller ROI size and the fact that the isolated CEST HIR were visually selected relative to surrounding signal intensity in CE-T1 tumor.
We demonstrated that a contrast in glioblastoma can be obtained by NOE mediated CEST imaging at 7T in terms of structure and extent of peritumoral hyperintensities and isolated CEST HIR that cannot be acquired with conventional CE-T1 and T2-weighted images.
As a principle finding of this study, we proofed, that the NOE-effects in glioblastoma CE-T1 tumors, as well as in the tumor necrosis and the surrounding PTCH within T2 edema margins are decreased in comparison to CLNAWM. This is in agreement with previously published studies
A lowered protein concentration might explain the NOE drop since water content in glioblastoma is supposed to be higher and extravasated serum proteins are reported to be lower than in healthy tissue
Since NOE mediated CEST imaging provides additional information compared to standard MR sequences, there are numerous possible clinical applications that need to be evaluated. For biopsy guidance, the commonly used CE-T1 images lack specificity, because they only visualize the extravasation of contrast agent due to a disrupted blood brain barrier. Subareas of different signal intensity within glioblastoma on NOE mediated CEST may therefore contribute to identify tumor parts of different malignity by adding information about protein concentration or protein folding.
Since we detected seven out of eight tumor satellites on MTRasym images as hyperintense, an additional use of NOE mediated CEST as endogenous contrast might increase sensitivity for the identification of tumor satellites. However, as CSF in brain sulci and ventricles as well as blood vessels also display hyperintense on MTRasym, specificity of NOE mediated CEST is limited and requires comparison to anatomic sequences.
Another major problem in glioblastoma imaging within daily clinical decision making is that it is not possible to differentiate a T2-signal increase caused by tumor infiltration from a non-specific cause of T2-signal increase (e.g. edema, radiation effects, decreased corticosteroid dosing, seizures, postoperative changes)
Other CEST studies on high grade glioma patients were performed by Wen et al.
To evaluate therapeutic response effects of chemotherapy with temozolomide in mice with human glioblastoma, Sagiyama et al. recently performed asymmetry based APT imaging on a 7T small animal MR system
Generally, when performing CEST asymmetry approaches, it has to be verified that the implicit assumption of competitive CEST effects is valid. At +3.5 ppm amide proton transfer (APT) of proteins was reported, while exchange relayed NOE occurs in the range from −2 to −5 ppm
Finally, the clinical benefit of the newly introduced NOE mediated CEST contrast still needs to be proven within larger patient collectives, including follow up examinations and bioptical correlations. Ultimately, beyond MTRasym approaches, sophisticated fitting models performed on Z-spectra with additional frequency offsets may help to separate competitive CEST effects yielding a metabolite specific CEST contrast.