5T‐Based Glutamate Chemical Exchange Saturation Transfer Imaging for Adult Diffuse Glioma Stratification

ABSTRACT

Background

Glutamate chemical exchange saturation transfer (GLU‐CEST) is a non‐invasive in vivo approach for glutamate detection, but its performance for glioma evaluation at 5T remains incompletely defined.

Purpose

To investigate factors influencing 5T GLU‐CEST and its diagnostic value in glioma stratification.

Study Type

Prospective.

Phantom and Population

Five phantom series (pH 6.2–7.4; glutamate 8–20 mM) and 40 adult‐type diffuse glioma patients (49.00 ± 12.26 years; 23 males).

Fieldstrength/Sequence

5 T, fast spin‐echo GLU‐CEST and amide proton transfer (APT)‐CEST.

Assessment

Phantoms and patients underwent MRI scans; GLU‐CEST, APT‐CEST, and normalized values (ΔGLU‐CEST, ΔAPT‐CEST) were quantified. Gliomas were graded by WHO 2–4; IDH status was determined.

Statistical Tests

Spearman correlation, Kruskal–Wallis tests, Mann–Whitney U tests, and weighted DeLong tests; two‐sided p < 0.05 was significant.

Results

GLU‐CEST signals positively correlated with glutamate and negatively with pH (ρ ≥ 0.893). Significant differences were found in all effects between WHO Grade 2 and 4 gliomas (ΔGLU‐CEST: 2.006 [1.143, 2.799] vs. 4.365 [2.974, 5.299]; GLU‐CEST: 7.424 [6.679, 7.649] vs. 10.155 [8.098, 11.550]; ΔAPT‐CEST: 1.918 [1.258, 2.461] vs. 3.386 [2.682, 4.805]; APT‐CEST: 1.961 [1.425, 2.715] vs. 3.333 [2.478, 4.632]), whereas only ΔGLU‐CEST and GLU‐CEST exhibited significant disparities between Grade 3 and 4 gliomas (ΔGLU‐CEST: 2.171 [1.895, 2.862] vs. 4.365 [2.974, 5.299]; GLU‐CEST: 7.102 [6.475, 7.259] vs. 10.155 [8.098, 11.550]). In the solid tumor region, all effects demonstrated significant differences between IDH‐mutant and IDH wild‐type gliomas (ΔGLU‐CEST: 2.111 [1.614, 3.110] vs. 4.333 [2.964, 5.405]; GLU‐CEST: 7.259 [6.577, 7.726] vs. 10.291 [8.097, 11.634]; ΔAPT‐CEST: 2.017 [1.355, 2.718] vs. 3.235 [2.670, 4.735]; APT‐CEST: 2.122 [1.680, 2.889] vs. 3.270 [2.450, 4.262]), whereas GLU‐CEST outperformed APT‐CEST and ΔAPT‐CEST in diagnostic efficacy (AUC difference = 0.073 and 0.101).

Data Conclusion

5 T GLU‐CEST is capable of differentiating between grade 2 and grade 4, as well as grade 3 and grade 4 adult diffuse gliomas, and demonstrates superior performance to APT‐CEST in the classification of IDH status.

Evidence Level

2.

Technical Efficacy

Stage 2.

Plain Language Summary

Gliomas are the most common type of brain tumor and remain difficult to treat. Glutamate accumulates at high levels in gliomas and is linked to tumor growth and treatment resistance. This study tested a new magnetic resonance imaging technique called GLU‐CEST at 5 Tesla, which is correlated with glutamate and pH in tissue. Using brain scans from 40 patients, the researchers found that GLU‐CEST successfully distinguished between different tumor grades and between tumors with or without IDH mutations, a key genetic marker. The technique also performed better than an existing method (APT‐CEST) for classifying tumor subtypes. These findings suggest that 5T GLU‐CEST may help doctors better characterize gliomas without invasive procedures.

1. Introduction

Gliomas, the most prevalent and primary cerebral tumors, exhibit infiltrative growth patterns, and intrinsic resistance to therapy, resulting in poor survival rates [1]. The updated World Health Organization (WHO) central nerve tumor classification combines histopathological features with genetic alterations including Isocitrate dehydrogenase (IDH) mutations and 1p/19q codeletion to refine tumor subtyping [2]. Despite the application of chemo‐ and radio‐therapy, as well as multiple novel types of treatment, including anti‐vascularization, immunotherapy, and tumor treating fields, the clinical outcomes for gliomas remain suboptimal, which may be attributed to the metabolic reprogramming [3, 4, 5]. This adaptive mechanism was shown to allow tumor cells to preferentially utilize metabolites to sustain uncontrolled proliferation and invasion while evading treatment‐induced stress [6].

Among all the metabolites involved in metabolic reprogramming, glutamate represents an important role in glioma pathobiology [7]. Malignant glioma cells disrupt cerebral glutamate homeostasis, accumulating extracellular glutamate concentrations up to 500‐fold higher than physiological levels [8]. This pathological buildup not only promotes excitotoxic neuronal death to facilitate tumor expansion but also establishes an immunosuppressive microenvironment [9]. The visualization of glutamate metabolism in glioma is critical for understanding metabolic behavior and developing targeted therapies, yet conventional imaging modalities, including proton magnetic resonance spectroscopy (MRS), face inherent limitations in spatially resolving glutamate dynamics due to low signal‐to‐noise ratios and spectral contamination from glutamine [10].

Chemical exchange saturation transfer (CEST) MRI provides a promising solution by exploiting the exchange kinetics of labile protons on endogenous metabolites [11]. As the most popular clinical use of this technique, amide proton transfer (APT)‐CEST imaging detects tumor‐associated proteins via amide proton resonance at 3.5 ppm, serving as a surrogate for cellular density and proliferation [12]. However, direct quantification of glutamate using APT‐CEST is unachievable at clinical field strengths (≤ 3T): its amine protons resonate at 3.0 ppm, where rapid proton exchange (~2 kHz) and macromolecular signal overlap severely compromise detectability [13].

The recent availability of 5T scanners overcomes these barriers by enhancing chemical shift dispersion to improve spectral separation between glutamate and confounding metabolites, while the increased resonance frequency optimizes saturation efficiency for intermediate exchange‐rate protons, thereby enabling glutamate‐specific CEST (GLU‐CEST) imaging [14]. Recent work by Zhou et al. has demonstrated that 5T GLU‐CEST imaging can achieve reproducible, high‐quality results, enabling noninvasive glutamate visualization [14]. However, despite these advances, the quantification accuracy of glutamate detection via 5T GLU‐CEST and its diagnostic efficacy specifically within glioma remain incompletely unexplored [14, 15].

This study aimed to investigate factors influencing GLU‐CEST using phantom models, and evaluate the accuracy of GLU‐CEST in glioma stratification through correlation with WHO histological grades and molecular subtypes, with a comparison of APT‐CEST.

2. Materials and Methods

This prospective and single‐center study was approved by our Institutional Review Board. All patients provided written informed consent. The overall flow diagram of the work is illustrated in Figure 1.

FIGURE 1.

The overall flow diagram of the study.

2.1. Phantom Sample Preparation

Glutamate was prepared at six concentrations (8 to 20 mM, increasing in 2 mM steps) in a phosphate‐based sodium‐potassium (PBS) buffer (titrated to pH 7.40 ± 0.02 using NaOH and HCl). Four additional sets of PBS solutions were prepared with different glutamate concentrations (5, 10, 15, and 20 mM), each set comprising seven samples with pH values systematically titrated from 6.2 to 7.4 at 0.2 pH unit intervals. All prepared solutions were subsequently transferred into 50 mL tubes, which were separately placed in five 11.4 cm‐diameter flasks containing 2% agarose solution. The phantom was solidified at 22°C ± 1°C before experiments.

2.2. Study Population of Patients

The inclusion criteria for patients were as follows: (1) suspicion of adult diffuse glioma based on preoperative MR imaging, (2) absence of intra‐tumoral hemorrhage, and (3) no prior cranial treatment. The exclusion criteria include: (1) MRI contraindications or severe motion artifacts, (2) reluctance to take surgery, (3) the interval between their MR scans and pathological examination exceeded 2 weeks, and (4) non‐adult diffuse glioma pathologically.

2.3. MR Imaging Acquisition

The images of phantoms and participants were obtained using a 5T MRI scanner (uMR Jupiter, United Imaging Healthcare, Shanghai, China) with a 48‐channel transceiver head coil. The scanning sequence for suspected glioma included axial and sagittal T1‐weighted fluid‐attenuated inversion recovery imaging (T1‐FLAIR), axial T2‐weighted imaging (T2WI), T2‐weighted FLAIR with fat suppression (T2‐FLAIR‐FS), and diffusion‐weighted imaging (DWI) sequences. The parameters for the above‐mentioned sequences were listed in Table S1. Based on the images obtained, the axial slice with the maximum tumor area was selected for CEST imaging.

Prior to CEST imaging, active shimming was performed, and the Water Saturation Shift Referencing (WASSR) method was used for B0 correction [16]. For the 2D Glu‐CEST imaging, the parameters were as follows: Repetition Time (TR) = 6000 ms; Echo Time (TE) = 6.6 ms; FOV = 220 × 220 mm²; acquisition matrix = 128 × 128; reconstruction matrix = 256 × 256; slice thickness = 10 mm; excitation flip angle = 90°; refocusing flip angle = 120°; band‐width = 500 Hz; shots per slice = 1; averages = 1. For the acquisition of saturated image Ssat, a rectangular‐shaped continuous wave radio frequency (RF) saturation pulse was applied, with a saturation intensity (B1rms) of 3 μT and a saturation duration of 1000 ms. A fat‐suppressed 2D single‐shot fast spin echo (SSH‐FSE) with an echo train length of 64 was used for acquisition. A B0Map was acquired utilizing a dual‐echo gradient echo (GRE) sequence with the same geometric parameters as CEST and TR = 7500 ms, TE = 2.76/5.52 ms, flip angle = 10°, band‐width = 700 Hz/Pixel. Glu‐CEST data were acquired at ±3 ppm, ±3.4 ppm, and ±2.6 ppm, with ±3 ppm acquired twice to increase Signal‐to‐Noise Ratio. A limited frequency offset range of ±0.4 ppm around the specific CEST peaks was sampled to prioritize scan efficiency for clinical feasibility [17].

The B1Map was generated employing a rapid B1Map method referenced in the literature [18]. Its geometric parameters were consistent with CEST, and nominal flip angle = 60° of the pre‐conditional pulse, bandwidth = 2500 Hz/Pixel. Additionally, a voxel‐wise B1 inhomogeneity correction was performed using piecewise linear interpolation of the signals acquired at different B1 strengths (4, 3, and 2.5 μT for Glu‐CEST and 3, 2, 1.5 μT for APT‐CEST) to derive the corrected signal at the target B1 strength [19]. The total acquisition time for each B1 value was approximately 11 s. Each acquisition time, comprising B0, B1, and S0, Ssat images, took 3 min 8 s.

For the 2D APT‐CEST imaging, the key parameters were listed as follows: TR = 6000 ms; TE = 6.7 ms; FOV = 220 × 220 mm²; acquisition matrix = 128 × 128; reconstruction matrix = 256 × 256; slice thickness = 5 mm; excitation flip angle = 90°; refocusing flip angle = 120°; band‐width = 500 Hz; shots per slice = 1; averages = 1. The same RF saturation pulse as in Glu‐CEST was used, but with a saturation intensity (B1rms) of 2 μT and a saturation duration of 2000 ms. APT‐CEST data were acquired at ±3.9, ±3.5, and ±3.1 ppm, with ±3.5 ppm acquired twice. The acquisition of B0, B1, and S0, Ssat images took about 3 min and 8 s in total.

Sagittal contrast‐enhanced T1WI (CE‐T1WI) was performed 5 min after intravenous administration of gadopentetate dimeglumine (Beilu, Beijing, China) at a dose of 0.1 mmol/kg. The parameters were listed in Table S1.

2.4. Imaging Post‐Processing

Imaging data were analyzed using the Medical Image Post‐processing Software uOmnispace.MR R002 (United Imaging Healthcare, Shanghai, China), which automatically generated GLU‐CEST and APT‐CEST maps.

The reconstruction process includes B0 and B1 corrections, and the calculation formulas for Glu‐CEST and APT‐CEST are as follows:

$$\text{CESTasym}\left(3\mathrm{ppm}\right)=\frac{\mathit{\text{Ssat}}\left(-3\mathrm{ppm}\right)-\mathit{\text{Ssat}}\left(3\mathrm{ppm}\right)}{\mathit{\text{Ssat}}\left(-3\mathrm{ppm}\right)}$$

$$\text{CESTasym}\left(3.5\mathrm{ppm}\right)=\frac{\mathit{\text{Ssat}}\left(-3.5\mathrm{ppm}\right)-\mathit{\text{Ssat}}\left(3.5\mathrm{ppm}\right)}{S0}$$

where 3 and 3.5 ppm are the chemical shifts of the glutamate and amide proton to free water, respectively. Asymmetric analysis of each voxel gives the final CEST quantitative map.

Image co‐registration and region of interest (ROI) delineation were performed using ITK‐SNAP software (version 3.8.0). Based on GLU‐CEST imaging, ROIs with an area of 200 mm² were placed at the center of each phantom tube. For patients, axial post‐processed GLU‐CEST and APT‐CEST images were rigidly registered to anatomical images, primarily axial T2‐FLAIR and T1CE images.

All tumor ROIs were drawn independently by two neuroradiologists (X.L. and Y.L., with 5 and 11 years of experience, respectively) who were blinded to clinical and pathological information.

ROIs were manually delineated using T2WI, T2‐FLAIR, and T1CE images as anatomical references. The ROIs included solid tumor region, peritumoral edema, and contralateral normal‐appearing white matter (CNAWM).

1. Solid tumor region: For gliomas with enhancement, ROIs were drawn around the contrast‐enhancing area (EA) on T1CE; for non‐enhanced gliomas, the solid tumor region is characterized by mild T2‐FLAIR hyperintensity, involvement of gray matter, eccentric growth beyond normal anatomical boundaries, localized tissue expansion, and associated mass effect [20, 21, 22]. Large cystic, necrotic, hemorrhagic, or vascular structures were excluded.

2. Edema: Edema was defined as FLAIR hyperintensity in white matter [23].

3. Contralateral normal‐appearing white matter (CNAWM): For each patient, six circular ROIs (~50 mm² each) were placed within the CNAWM. This region was used as the reference for signal normalization.

Mean GLU‐CEST and APT‐CEST signal intensities were extracted from each ROI. Normalized GLU‐CEST/APT‐CEST values were calculated by subtracting the CNAWM value from the tumor value (tumor ROI − CNAWM) and were respectively denoted as ΔGLU‐CEST and ΔAPT‐CEST.

2.5. Clinical and Pathological Data Collection

All clinical data were collected from medical records and patient history. All enrolled patients underwent surgical resection. Tumor specimens were graded by two pathologists independently (with 8 and 15 years of working experience, respectively) according to the WHO CNS5, resolving discrepancies by consensus. IDH mutation status (codons 132 of IDH1 and 140/172 of IDH2) and TERT promoter (pTERT) mutations (C228T and C250T) were assessed by Sanger sequencing. The status of 1p/19q co‐deletion was determined using fluorescence in situ hybridization.

2.6. Statistical Analysis

Statistical analyses were carried out with SPSS statistical software (version 26.0) and Python (version 3.8). To assess reproducibility of ROIs, intraclass correlation coefficients (ICCs) were calculated for GLU‐CEST and APT‐CEST values extracted from each ROI pair. The B0 offset values are reported as mean ± SD (ppm) within the ROI. Spearman correlation analysis was performed to assess the relationship between glutamate concentration/pH and GLU‐CEST signal intensity. Friedman M with Nemenyi tests and Bonferroni correlations were used to compare regional GLU‐CEST and APT‐CEST values within individuals. The Kruskal–Wallis tests with post hoc pairwise comparisons (Dunn's test with Bonferroni correction) were used to reveal statistically significant differences in CEST effects across the three WHO grades. Group comparisons were performed using the Mann–Whitney U tests. The area under the curve (AUC) was calculated using the bootstrap resampling method to assess the diagnostic performance. Given the imbalance in sample size between subgroups (IDH wild‐type: 25 cases; IDH mutant: 15 cases), the weighted DeLong tests with Benjamini–Hochberg corrections were employed to assess whether the differences in AUC values were statistically significant. A two‐sided p value < 0.05 was considered statistically significant.

It was hypothesized that ΔGLU‐CEST in the solid tumor region would differ by IDH status and WHO grade. Thus, these two comparisons were designated as primary endpoints. All other analyses are exploratory.

3. Results

3.1. Phantom Sample Analysis

In phantom studies, GLU‐CEST signals demonstrated dependence on glutamate concentration and pH levels. At pH 7.4, GLU‐CEST values showed a concentration‐dependent increase from 8 to 20 mM glutamate (0.350% to 1.033%, Figure 2), with a strong positive correlation (ρ = 0.893). More notably, a significant increase in GLU‐CEST signal intensity with decreasing pH levels from 7.4 to 6.2 was observed at fixed glutamate concentrations (5–20 mM, Figure 3). This inverse relationship was highly consistent across all tested glutamate concentrations (all ρ = −1.000).

FIGURE 2.

The results of phantom sample analysis at pH 7.4 and varied glutamate concentration (8 mM to 20 mM, increasing in 2 mM steps).

FIGURE 3.

The results of phantom sample analysis at pH 6.2–7.4 and varied glutamate concentration (5 mM, 10 mM, 15 mM, and 20 mM).

3.2. Patient Characteristics

Between July 2024 and March 2025, a total of 52 consecutive patients were initially enrolled. Twelve patients were excluded in accordance with the exclusion criteria, and finally 40 patients (mean age: 49.00 ± 12.26 years, 23 males) with a confirmed diagnosis of adult‐type diffuse glioma were included in this study. Tumors were most commonly located in the frontal lobe (n = 22, 55.0%), and 4 (10%) patients had a history of epilepsy. According to 2021 WHO classification, 9 (22.5%) patients were classified as WHO grade 2, 5 (12.5%) as grade 3, and 26 (65.0%) as WHO grade 4. Based on integrated histopathological and molecular diagnoses, the cohort included 25 (62.5%) glioblastomas, 8 (20.0%) oligodendrogliomas, and 7 (17.5%) astrocytomas.

IDH1/2 and pTERT (C250T/C228T) mutation status was determined in all patients: 15 (37.5%) had IDH mutations and 25 (62.5%) were IDH wild‐type; 26 (65.0%) had pTERT mutations, and 14 (35.0%) were wild‐type. Additionally, among the 15 IDH‐mutated gliomas, 8 (53.3%) had 1p19q co‐deletions and 7 (46.7%) had intact 1p19q. Detailed clinical and pathological information of the cohort is listed in Table 1.

TABLE 1.

Clinicopathological characteristics of patients with adult diffuse glioma.

Parameter	Value
No. of patients	40
Sex (male:female)	23:17
Age (years)	49.00 ± 12.26
Epilepsy
Yes	4 (10%)
No	36 (90%)
Location
Frontal lobe	22 (55.0%)
Temporal lobe	8 (20.0%)
Occipital lobe	5 (12.5%)
Thalamus	3 (7.5%)
Parietal lobe	2 (5.0%)
Integrated diagnosis
Oligodendroglioma	8 (20.0%)
Astrocytoma	7 (17.5%)
Glioblastoma	25 (62.5%)
WHO grade
2	9 (22.5%)
3	5 (12.5%)
4	26 (65.0%)
High/low‐grade
Low‐grade glioma	9 (22.5%)
High‐grade glioma	31 (77.5%)
IDH1/2
Mutated	15 (37.5%)
Wild	25 (62.5%)
TERT promoter C250/C228
Mutated	26 (65.0%)
Wild	14 (35.0%)
1p19q (among the 15 IDH‐mutated gliomas)
Co‐deleted	8 (53.3%)
Intact	7 (46.7%)

3.3. ROI Reproducibility and B0 Field Uniformity

The inter‐observer agreement results for ROI reproducibility are detailed in Text S1. All intraclass correlation coefficients values exceeded 0.835.

After WASSR correction, the B0 shifts within the ROI were 0.108 ± 0.077 ppm (range: 0.026 to 0.285 ppm) for GLU‐CEST and 0.108 ± 0.055 ppm (range: 0.016 to 0.292 ppm) for APT‐CEST.

3.4. GLU‐CEST and APT‐CEST Signal Characteristics Across Tumor, Edema, and CNAWM

As depicted in Figure 4 and Table S2, GLU‐CEST and APT‐CEST imaging revealed significant differences in quantitative signal values across the solid tumor region, peritumoral edema, and CNAWM.

FIGURE 4.

GLU‐CEST and APT‐CEST contrast across tumor, edema, and CNAWM.

In the CNAWM, GLU‐CEST, and APT‐CEST signals were the lowest compared with those in the solid tumor and peritumoral edema regions (APT‐CEST: solid tumor 2.889 [2.130, 3.836], peritumoral edema 1.590 [1.032, 2.319], CNAWM −0.027 [−0.329, 0.339]; GLU‐CEST: solid tumor 8.317 [7.348, 10.907], peritumoral edema 7.802 [6.869, 8.843], CNAWM 5.592 [5.064, 6.017]). Specifically, the GLU‐CEST and APT‐CEST values in the solid tumor region were markedly elevated relative to their corresponding values in the CNAWM. As for the peritumoral edematous region, GLU‐CEST and APT‐CEST signals were significantly lower than in the tumor region but remained elevated compared to CNAWM (Table S2 and Figure 4).

3.5. The Relationship Between GLU‐CEST/APT‐CEST Values and WHO Grade

Figure 5 presents the anatomical and CEST images for a Glioblastoma (WHO Grade 4) and two Astrocytomas (WHO Grade 2 and 3), respectively.

FIGURE 5.

MR images of three glioma patients. (A) A representative case of WHO Grade 2 glioma (astrocytoma) in a 43‐year‐old female patient. (B) A representative case of WHO Grade 3 (astrocytoma) glioma in a 35‐year‐old male patient. (C) A representative case of WHO Grade 4 glioma (glioblastoma) in a 42‐year‐old female patient. The T2‐FLAIR images show the ROIs (red: Solid tumor regions; blue: Peritumoral edema regions). Both APT‐CEST and GLU‐CEST images show a pronounced increase in the glioblastoma, compared with normal‐appearing brain tissue. In the astrocytomas, APT‐CEST shows moderate or pronounced increase in the tumor. However, the tumor regions of the astrocytomas showed no significant or mild GLU‐CEST signal elevation and demonstrated marked intratumoral signal heterogeneity in the WHO Grade 2 glioma.

Table 2 presents the distribution of four CEST effects (ΔGLU‐CEST, GLU‐CEST, ΔAPT‐CEST, and APT‐CEST) in solid tumor regions among WHO Grade 2 (n = 9), grade 3 (n = 5), and grade 4 (n = 26) gliomas. The Kruskal‐Wallis test revealed statistically significant differences in all CEST effects across different WHO Grades (ΔGLU‐CEST: 2.006 [1.143, 2.799] in Grade 2, 2.171 [1.895, 2.862] in Grade 3, and 4.365 [2.974, 5.299] in Grade 4; GLU‐CEST: 7.424 [6.679, 7.649] in Grade 2, 7.102 [6.475, 7.259] in Grade 3, and 10.155 [8.098, 11.550] in Grade 4; ΔAPT‐CEST: 1.918 [1.258, 2.461] in Grade 2, 2.220 [1.898, 2.226] in Grade 3, and 3.386 [2.682, 4.805] in Grade 4; APT‐CEST: 1.961 [1.425, 2.715] in Grade 2, 2.132 [2.122, 2.166] in Grade 3, and 3.333 [2.478, 4.632] in Grade 4).

TABLE 2.

Comparison of CEST effects in the solid tumor regions among gliomas of different WHO grades.

Effects	Grade 2 (n = 9)	Grade 3 (n = 5)	Grade 4 (n = 26)	χ 2 value	p
ΔGLU‐CEST	2.006 [1.143, 2.799]	2.171[1.895, 2.862]	4.365 [2.974, 5.299]	14.931	< 0.001
GLU‐CEST	7.424 [6.679, 7.649]	7.102 [6.475, 7.259]	10.155 [8.098, 11.550]	16.487	< 0.001
ΔAPT‐CEST	1.918 [1.258, 2.461]	2.220 [1.898, 2.226]	3.386 [2.682, 4.805]	13.067	0.002
APT‐CEST	1.961 [1.425, 2.715]	2.132 [2.122, 2.166]	3.333 [2.478, 4.632]	11.560	0.003

Post hoc pairwise comparisons via Dunn's test (Table 3) further delineated these differences: between Grade 2 and Grade 4, all four CEST effects exhibited significant disparities (ΔGLU‐CEST: 2.006 [1.143, 2.799] vs. 4.365 [2.974, 5.299]; GLU‐CEST: 7.424 [6.679, 7.649] vs. 10.155 [8.098, 11.550]; ΔAPT‐CEST: 1.918 [1.258, 2.461] vs. 3.386 [2.682, 4.805]; APT‐CEST: 1.961 [1.425, 2.715] vs. 3.333 [2.478, 4.632]). For Grade 3 vs. Grade 4, significant differences were observed in ΔGLU‐CEST (2.171 [1.895, 2.862] vs. 4.365 [2.974, 5.299]) and GLU‐CEST (7.102 [6.475, 7.259] vs. 10.155 [8.098, 11.550]), though no significant distinctions emerged for ΔAPT‐CEST or APT‐CEST (ΔAPT‐CEST: 2.220 [1.898, 2.226] vs. 3.386 [2.682, 4.805], p = 0.129; APT‐CEST: 2.132 [2.122, 2.166] vs. 3.333 [2.478, 4.632], p = 0.202). Notably, no significant differences in any CEST effect were detected between Grade 2 and Grade 3 (ΔGLU‐CEST: 2.006 [1.143, 2.799] vs. 2.171 [1.895, 2.862], p = 1.000; GLU‐CEST: 7.424 [6.679, 7.649] vs. 7.102 [6.475, 7.259], p = 1.000; ΔAPT‐CEST: 1.918 [1.258, 2.461] vs. 2.220 [1.898, 2.226], p = 1.000; APT‐CEST: 1.961 [1.425, 2.715] vs. 2.132 [2.122, 2.166], p = 1.000).

TABLE 3.

Post hoc multiple comparisons of CEST effects in solid tumor regions among different WHO grades of gliomas.

Comparison groups		Z value [95% CI]	Cliff's delta [95% CI]	Corrected p
Grade 2 vs. 3	ΔGLU‐CEST	−0.242	−0.200 [−0.527, 0.127]	1.000
	GLU‐CEST	0.211	0.156 [−0.171, 0.482]	1.000
	ΔAPT‐CEST	−0.549	−0.289 [−0.616, 0.038]	1.000
	APT‐CEST	−0.603	−0.378 [−0.704, −0.051]	1.000
Grade 2 vs. 4	ΔGLU‐CEST	−3.430	−0.752 [−0.974, −0.530]	0.002
	GLU‐CEST	−3.367	−0.778 [−1.000, −0.556]	0.002
	ΔAPT‐CEST	−3.345	−0.735 [−0.957, −0.513]	0.003
	APT‐CEST	−3.179	−0.684 [−0.906, −0.462]	0.004
Grade 3 vs. 4	ΔGLU‐CEST	−2.440	−0.738 [−1.000, −0.458]	0.044
	GLU‐CEST	−2.908	−0.800 [−1.000, −0.519]	0.011
	ΔAPT‐CEST	−2.023	−0.615 [−0.896, −0.335]	0.129
	APT‐CEST	−1.829	−0.585 [−0.865, −0.304]	0.202

Peritumoral edema was present in 67.5% of cases (27/40). Due to the limited sample size of WHO Grade 2 (n = 3) and Grade 3 (n = 2) gliomas with edema, we only performed an exploratory analysis, and the corresponding results are presented in Table S3.

3.6. The Relationship Between GLU‐CEST/APT‐CEST Contrast and Molecular Expression

Table 4 compares four CEST effects in solid tumor regions between IDH‐mutant (n = 15) and IDH wild‐type (n = 25) gliomas. All CEST effects were significantly lower in the IDH‐mutant group: ΔGLU‐CEST (2.111 [1.614, 3.110] vs. 4.333 [2.964, 5.405]; AUC = 0.835 [95% CI: 0.702, 0.942]), GLU‐CEST (7.259 [6.577, 7.726] vs. 10.291 [8.097, 11.634]; AUC = 0.864 [95% CI: 0.745, 0.959]), ΔAPT‐CEST (2.017 [1.355, 2.718] vs. 3.235 [2.67, 4.735]; AUC = 0.789 [95% CI: 0.608, 0.939]), and APT‐CEST (2.122 [1.68, 2.889] vs. 3.270 [2.45, 4.262]; AUC = 0.771 [95% CI: 0.583, 0.912]).

TABLE 4.

The comparison of CEST effects in solid tumor region between IDH‐mutant and IDH wild‐type gliomas.

Effects	IDH‐mutant (n = 15)	IDH wild‐type (n = 25)	Z value [95% CI]	p
ΔGLU‐CEST	2.111 [1.614, 3.110]	4.333 [2.964, 5.405]	−3.506 [0.634, 0.705]	< 0.001
GLU‐CEST	7.259 [6.577, 7.726]	10.291 [8.097, 11.634]	−3.813 [0.692, 0.764]	< 0.001
ΔAPT‐CEST	2.017 [1.355, 2.718]	3.235 [2.67, 4.735]	−3.031 [0.543, 0.614]	0.002
APT‐CEST	2.122 [1.68, 2.889]	3.270 [2.45, 4.262]	−2.836 [0.506, 0.577]	0.005

Weighted DeLong tests with Benjamini–Hochberg correction were performed to compare the diagnostic performance of 5T CEST parameters in the tumor solid region for differentiating IDH status (Table S4). Among all pairwise comparisons, GLU‐CEST vs. APT‐CEST exhibited the largest AUC difference (0.101) and statistical significance, followed by GLU‐CEST vs. ΔAPT‐CEST (AUC difference = 0.073) and ΔGLU‐CEST vs. APT‐CEST (AUC difference = 0.069). All other pairwise comparisons of AUC values did not reach statistical significance after correction (ΔGLU‐CEST vs. GLU‐CEST: AUC difference = −0.031, Corrected p = 0.178; ΔGLU‐CEST vs. APT‐CEST: AUC difference = 0.042, Corrected p = 0.101; ΔAPT‐CEST vs. APT‐CEST: AUC difference = 0.028, Corrected p = 0.178).

Table 5 presents CEST effects in edema region of IDH‐mutant (n = 6) and IDH wild‐type (n = 21) gliomas. Significant differences were observed for ΔGLU‐CEST (IDH‐mutant: 1.312 [−0.058, 1.933] vs. IDH wild‐type: 2.829 [1.620, 3.816]) and GLU‐CEST (IDH‐mutant: 6.156 [5.313, 6.801] vs. IDH wild‐type: 7.881 [7.594, 9.728]). In contrast, ΔAPT‐CEST (IDH‐mutant: 2.022 [1.087, 2.319] vs. IDH wild‐type: 1.657 [1.156, 2.052], p = 0.861) and APT‐CEST (IDH‐mutant: 1.971 [0.808, 2.513] vs. IDH wild‐type: 1.590 [1.191, 2.109], p = 1.000) showed no significant differences between the two groups.

TABLE 5.

The comparison of CEST effects in edema region between IDH‐mutant and IDH wild‐type gliomas.

Effects	IDH‐mutant (n = 6)	IDH wild‐type (n = 21)	Z value	p
ΔGLU‐CEST	1.312 [−0.058, 1.933]	2.829 [1.620, 3.816]	−2.158	0.031
GLU‐CEST	6.156 [5.313, 6.801]	7.881 [7.594, 9.728]	−2.800	0.005
ΔAPT‐CEST	2.022 [1.087, 2.319]	1.657 [1.156, 2.052]	−0.175	0.861
APT‐CEST	1.971 [0.808, 2.513]	1.590 [1.191, 2.109]	0.000	1.000

Tables S4 and S5 compare CEST effects in the solid and edema regions between TERT promoter mutant and wild‐type gliomas, respectively. No significant differences were observed across all CEST effects in either region (all p > 0.050).

4. Discussion

Glutamate plays a critical role in the initiation and malignant progression of gliomas by promoting proliferation, neuronal toxicity, and infiltration [24, 25]. Although the pathological role of glutamate in glioma progression has been well established and several therapeutic strategies targeting glutamate metabolism are under clinical investigation, in vivo imaging of glutamate in glioma patients remains limited [14]. Through phantom experiments, we found that within both pathological and physiological ranges, a lower pH or a higher glutamate concentration was associated with a stronger 5T GLU‐CEST signal. Notably, the signal amplification effect of an acidic environment was more pronounced at high glutamate concentrations. In vivo experiments at 5T demonstrated that GLU‐CEST enables differentiation of adult diffuse gliomas between WHO Grade 2 and Grade 4, and between Grade 3 and Grade 4, and exhibits superior performance over APT‐CEST in the determination of IDH status. However, neither GLU‐CEST nor APT‐CEST seems to be capable of differentiating pTERT status.

Through systematic aqueous phantom experiments at 5T, we characterized the dependence of GLU‐CEST contrast on glutamate concentration and pH. Based on our observation, pH exerted a stronger influence on GLU‐CEST signals than glutamate concentration itself. Under physio‐ and pathological‐relevant glutamate levels (8–20 mM, approximating brain tissue concentrations of ~10 mM [26]), elevated pH (7.4) minimized signal changes with varying glutamate. Conversely, at fixed glutamate concentrations, pH variations significantly modulated GLU‐CEST signals, particularly at higher glutamate levels (e.g., 20 mM). This occurs because acidic conditions accelerate chemical exchange rates between glutamate amine protons and bulk water protons, amplifying CEST effects [18]. Notably, glutamate concentration and pH are intrinsically coupled: increased glutamate lowers microenvironmental pH (e.g., via excitotoxic mechanisms), while reduced pH further enhances exchange kinetics. This interdependence precludes isolated quantification of glutamate concentration. Our findings align with Cai et al.'s 7T study demonstrating linear GLU‐CEST‐pH dependence within pH 6.0–7.4 [18]. Moreover, studies have demonstrated the influence of other factors, such as direct water saturation, semi‐solid magnetization transfer, T1‐weighting effects, and macromolecular background on GLU‐CEST signals [27, 28, 29]. Therefore, we propose that GLU‐CEST represents a composite biomarker of glutamate‐associated acidosis rather than a direct measure of glutamate concentration.

In this study, GLU‐CEST enabled differentiation of adult diffuse gliomas between WHO Grade 2 and Grade 4, and between Grade 3 and Grade 4. We also found that the GLU‐CEST signal in the peritumoral edema region of IDH‐wildtype gliomas was significantly higher than that of IDH‐mutant gliomas, whereas this difference was not statistically significant for APT‐CEST. These results indicate that GLU‐CEST is sensitive to both central and infiltrative metabolic activity in glioma. Given that APT‐CEST reflects increased mobile proteins as a secondary effect of tumor metabolic reprogramming and is also affected by pH, we propose that the changes of GLU‐CEST in the peritumoral edema area may be associated with early‐stage glutamate release. These findings are consistent with the glutamate‐mediated neurotoxicity model, in which glioma cells release excessive extracellular glutamate to induce excitotoxic neuronal death, thereby providing space and a permissive microenvironment for tumor infiltration [8, 30]. Our finding hints at the potential of GLU‐CEST to characterize the biological features of peritumoral edema. However, given the current sample size, further validation (e.g., larger cohorts, multi‐center replication) is required before drawing inferences about GLU‐CEST's utility in identifying the true extent of early tumor infiltration.

We also evaluated the association between GLU‐CEST signals and key molecular alterations in gliomas. ΔGLU‐CEST contrast was significantly lower in IDH‐mutant tumors, consistent with previous MRS‐ and mass spectrometry‐based studies [31]. This observation aligns with the known metabolic reprogramming induced by mutant IDH enzymes, which consume α‐ketoglutarate (α‐KG) to produce the oncometabolite 2‐hydroxyglutarate (2‐HG) [32]. To sustain this altered metabolic pathway activity, IDH‐mutant cells increasingly rely on glutaminolysis, converting glutamine to glutamate and subsequently to α‐KG, ultimately leading to a depletion of intracellular glutamate pools detectable by GLU‐CEST imaging [32]. No significant associations were observed between CEST metrics and pTERT status, which is primarily linked to telomere maintenance and cellular immortality [33]. These findings highlight the metabolic specificity of GLU‐CEST for IDH‐related alterations and its potential as a biomarker for molecular stratification.

Previous studies have suggested that elevated glutamate levels may increase neuronal excitability and contribute to glioma‐associated drug‐resistant epilepsy, with higher GLU‐CEST signals reported in patients presenting with seizures [15, 34]. But in our cohort, only 4 (10%) patients had a documented history of epilepsy, and no clear association between GLU‐CEST signal and seizure status could be established. This discrepancy may be attributed to differences in patient selection, tumor location, or cohort size. In particular, the predominance of frontal lobe tumors in our cohort—regions less commonly associated with epileptogenic activity compared to temporal lobe involvement—may partially explain the low seizure prevalence observed.

5. Limitations

This study was conducted using a single MR vendor/scanner and field strength. The small and imbalanced sample compromises generalizability. Findings beyond the primary endpoints should be interpreted as hypothesis‐generating. The use of different slice thicknesses and a single‐slice acquisition may introduce partial volume effects. Future studies will benefit from multi‐slice acquisitions with thinner slices (e.g., 3–5 mm) matching routine clinical protocols. Without T1 relaxation correction, glioma magnetization transfer ratio measurements may suffer T1‐related bias. Our phantom only modulated pH and glutamate, excluding macromolecular background signals, tissue relaxation properties, and other metabolites—factors of in vivo GLU‐CEST signals, which limits direct translation of phantom results to clinical brain imaging.

6. Conclusion

Although GLU‐CEST measurements at 5T are influenced by multiple factors, they can still serve as a reliable imaging biomarker for stratifying adult diffuse gliomas. In this study, 5T GLU‐CEST demonstrated the capability to differentiate between different WHO Grade adult diffuse gliomas, while showing superior performance over APT‐CEST in determining tumor IDH status.

Funding

National Natural Science Foundation of China (82281966); Explorers Program of Shanghai (24TS1410800).

Conflicts of Interest

The authors declare no conflicts of interest.

Supporting information

Data S1: jmri70340‐sup‐0001‐supinfo.docx.

Table S1: Scanning parameters of MRI routine protocols in this study.

Table S2: The differences in GLU‐CEST/APT‐CEST contrast among solid tumor, peritumoral edema, and contralateral normal apparent white matter.

Table S3: The comparison of CEST effects in the edema regions among gliomas of different WHO grades.

Table S4: Diagnostic performance analysis of 5T CEST parameters in tumor solid region in differentiating IDH status in gliomas.

Table S5:. The comparison of CEST effects in solid region between TERT promoter mutant and TERT promoter wild‐type gliomas.

Table S6:. The comparison of CEST effects in edema region between TERT promoter mutant and TERT promoter wild‐type gliomas.