Amide Proton Transfer Imaging for Differentiating Gliomas and Meningiomas and Assessing Tumor Infiltration

Summary

This study evaluates Amide Proton Transfer (APT) imaging combined with conventional magnetic resonance imaging (MRI) to differentiate meningiomas from low- and high-grade gliomas and assess imaging features related to tumor infiltration, demonstrating its potential to improve preoperative classification and provide imaging-based information that may inform surgical planning.

Abstract

Differentiating gliomas from meningiomas and assessing imaging features related to glioma infiltration remain challenging with conventional magnetic resonance imaging (MRI). Amide Proton Transfer (APT) imaging, a chemical exchange saturation transfer technique, can detect changes in tissue protein content and pH, offering potential functional insights into brain tumors. This retrospective study evaluated APT imaging combined with conventional MRI in 105 patients with pathologically confirmed brain tumors (55 meningiomas, 20 low-grade gliomas [LGGs], and 30 high-grade gliomas [HGGs]). APT-derived mean signal intensity (APT_mean) and the ratios of APT abnormal area to T2 hyperintense area (RAPT/T2) and to contrast-enhancing area (RAPT/E) were measured and compared among groups. APTmean was significantly higher in HGGs (median 4.60%) than in LGGs (2.95%) and meningiomas (2.70%, P < 0.001). RAPT/T2 and RAPT/E were also markedly elevated in HGGs. For distinguishing HGGs from meningiomas, APTmean achieved excellent diagnostic performance (area under the curve [AUC] = 0.969), compared with RAPT/E (0.836) and RAPT/T2 (0.800). For differentiating LGGs from meningiomas, RAPT/E (AUC = 0.781) and RAPT/T2 (AUC = 0.734) showed good performance, whereas APT_mean did not differ significantly. For glioma grading (HGG vs. LGG), APTmean yielded an AUC of 0.916, superior to contrast enhancement degree (0.823). These findings indicate that APT imaging effectively distinguishes meningiomas from gliomas of different grades and provides valuable imaging-based information for inferring the extent of tumor infiltration. The combination of APTmean and spatial extent ratios provides a practical imaging tool for improving preoperative diagnosis and supporting surgical planning, particularly in challenging atypical cases.

Introduction

Intracranial tumors are among the most common disorders of the nervous system. Their incidence is increasing globally, posing a serious threat to human health¹. Based on histopathological origin, intracranial tumors can be classified into two major categories: primary and secondary (metastatic). Among primary intracranial tumors, meningioma and glioma represent the two most prevalent types with distinctly different origins and biological behaviors². Glioma, originating from glial cells, is the most common malignant intracranial tumor, characterized by infiltrative growth, ill-defined borders, a high tendency for recurrence, and a poor prognosis^3,4. The World Health Organization (WHO) classifies gliomas into grades I-IV based on their malignancy. Among these, high-grade gliomas (HGGs, WHO grades III-IV) are highly invasive and are associated with a short patient survival period^5,6. In contrast, meningioma originates from arachnoid cap cells. The vast majority are benign (WHO grade I), grow slowly, and have clear boundaries with a distinct demarcation from the surrounding brain tissue. Following complete surgical resection, patients typically have a favorable prognosis^7,8. Therefore, accurate preoperative imaging differentiation between these two tumor types, along with assessment of imaging features related to the extent of glioma infiltration, holds critical clinical significance. It is essential for formulating individualized surgical plans, determining safe resection margins, and ultimately improving patient outcomes.

Currently, magnetic resonance imaging (MRI) has become a widely used modality for preoperative diagnosis and assessment of intracranial tumors due to its superior soft tissue resolution, multi-parameter imaging capabilities, and the advantage of being non-ionizing^9,10. Conventional MRI sequences, such as T1-weighted imaging (T1WI), T2-weighted imaging (T2WI), fluid-attenuated inversion recovery (FLAIR), and contrast-enhanced T1-weighted imaging (CE-T1WI), can clearly reveal the tumor's location, size, morphology, internal structure, and its relationship with adjacent tissues^11,12. Typical meningiomas appear on MRI as well-defined extra-axial masses, exhibiting iso- or slightly hypointense signals on T1WI, iso- or slightly hyperintense signals on T2WI, with significant and homogeneous enhancement post-contrast, often accompanied by a "dural tail sign"¹³. In contrast, high-grade gliomas (HGGs) typically present as irregularly shaped, poorly defined masses. They are often surrounded by extensive peritumoral edema visible on T2WI/FLAIR sequences and demonstrate heterogeneous ring-like or garland-like enhancement after contrast administration^14,15.

However, conventional MRI still faces numerous challenges in clinical practice. Regarding differential diagnosis, some cases present with atypical features. For instance, certain HGGs may exhibit a relatively well-defined, homogeneously enhancing "pseudocapsule," making them difficult to distinguish from meningiomas. Conversely, some high-grade, cellular atypical or anaplastic meningiomas (WHO grades II-III), due to their invasive growth patterns, may also radiologically mimic gliomas¹⁶. Furthermore, conventional MRI has limited ability to accurately assess the presumed extent of glioma infiltration. The hyperintense region observed on T2WI/FLAIR sequences has been pathologically confirmed to be an admixture of tumor cell infiltration, vasogenic edema, and reactive gliosis. Conventional MRI is unable to effectively distinguish these components¹⁷. Similarly, the enhancing region visualized on CE-T1WI primarily represents the tumor core with a disrupted blood-brain barrier. A significant number of proliferative tumor cells have already infiltrated beyond this enhancing region, into and sometimes beyond the peritumoral T2 hyperintense area within the normal-appearing brain parenchyma^18,19. This mismatch between the radiological and pathological boundaries is a key contributor to the high post-surgical recurrence rate of gliomas. Consequently, there is an urgent clinical need for novel molecular imaging tools that can more directly and sensitively reflect tumor cellularity and metabolic activity.

Amide Proton Transfer (APT) imaging is an emerging molecular magnetic resonance imaging technique. It operates based on the Chemical Exchange Saturation Transfer (CEST) mechanism^20,21. By selectively saturating the amide protons (-NH) on the backbone of mobile proteins and peptides within tissues and transferring this magnetization to surrounding water molecules, it indirectly detects changes in tissue protein content and intracellular/extracellular pH^22,23. In the tumor microenvironment, malignant cells typically exhibit intense proliferative activity, leading to a significant increase in protein anabolism²⁴. Concurrently, enhanced glycolysis (the Warburg effect) in tumor tissues often results in acidification of the extracellular microenvironment²⁵. These pathophysiological alterations generally cause APT signals to appear characteristically hyperintense in high-grade malignancies. Furthermore, APT has demonstrated potential value in differentiating between glioma recurrence and radiation necrosis, offering complementary information to conventional MRI and other functional imaging techniques²⁶. This is because recurrent tumor tissue, characterized by high cellular density and active metabolism, exhibits high APT signals. In contrast, radiation necrosis, composed primarily of cellular debris and inflammatory reaction, presents with low APT signals. These findings have been corroborated by clinical studies demonstrating the superiority of APT over conventional MRI in distinguishing tumor recurrence from treatment-related changes²⁷.

However, current research predominantly focuses on glioma grading or differentiation from other intracranial pathologies, such as metastases or lymphoma^28,29,30. Systematic studies specifically investigating the utility of APT imaging for distinguishing meningiomas from gliomas of varying grades, particularly those encompassing both low-grade and high-grade gliomas, remain relatively limited. Furthermore, whether the area of APT signal abnormality may serve as an imaging surrogate that better correlates with the extent of cellular infiltration in gliomas, especially high-grade gliomas (HGGs), warrants further in-depth investigation.

Clinically, APT imaging may be valuable for preoperative differentiation of HGGs from meningiomas and for assessing imaging features related to glioma infiltration, especially when conventional MRI is equivocal. However, routine implementation faces practical challenges: prolonged acquisition time increases motion risk; signal quantification is sensitive to field inhomogeneities and post-processing algorithms, requiring standardized protocols; and APT metrics should complement, not replace, conventional MRI due to potential overlap in some LGGs and benign meningiomas. For adoption, we recommend starting with challenging cases and gradually integrating into routine practice once local expertise and quality assurance are established. Ultimately, APT imaging offers unique biological information but is best used judiciously within a multiparametric MRI framework.

Prior work by Zhang et al.³¹ has preliminarily explored the utility of APT imaging in distinguishing meningiomas from gliomas, proposing the RAPT/T2 and RAPT/E ratios as potential markers of tumor infiltration. However, that study was limited by a relatively small sample size (50 patients) and did not systematically evaluate the diagnostic performance of APT parameters compared with conventional MRI features across all clinically relevant pairwise distinctions, using comprehensive receiver operating characteristic (ROC) curve analyses. Furthermore, the clinical implications of APT-defined spatial extent abnormalities for surgical and radiotherapy planning were not thoroughly discussed. Based on this rationale, this study aims to retrospectively analyze APT and conventional MRI data from patients with pathologically confirmed meningiomas and gliomas of different grades. We systematically compare APT parameters between these two tumor types to evaluate the diagnostic efficacy of APT imaging for the preoperative differentiation of meningiomas from gliomas. Concurrently, by comparing the spatial extent of the APT abnormality region with that of the T2WI hyperintensity and CE-T1WI enhancement areas, this study preliminarily explores the potential application value of APT imaging as a radiological tool for inferring the infiltrative growth boundary of tumors. The objective is to provide clinicians with imaging-derived information that may strengthen the foundation for surgical planning, although such inferences await pathological validation.

Protocol

This study protocol adhered to the ethical principles outlined in the Declaration of Helsinki and was approved by the Shenzhen Bao'an District Songgang People's Hospital ethics review board (Approval number: IRB-YJ-2025-045). Given its retrospective design, the study involved only the analysis of archived clinical and imaging data. All patient-identifiable information was strictly anonymized.

Research design and subjects

This study is a retrospective diagnostic analysis. We retrospectively collected clinical and imaging data from brain tumor patients admitted to the Neurosurgical Oncology Departments of Songgang People's Hospital, Bao'an District, Shenzhen, and Shenzhen Second People's Hospital between January 2024 and June 2025. All patients had pathologically confirmed diagnoses post-surgery. A total of 105 patients were finally included, all of whom underwent both conventional MRI and APT imaging examinations preoperatively. According to the 2021 WHO Classification of Tumors of the Central Nervous System³², the patients were divided into three groups: Meningioma Group: 55 cases, pathologically confirmed as WHO grades I-II. Low-Grade Glioma (LGG) Group: 20 cases, pathologically confirmed as WHO grades I-II. HGG Group: 30 cases, pathologically confirmed as WHO grades III-IV. All patients underwent both conventional MRI and APT imaging examinations preoperatively. The detailed study flowchart is presented in Figure 1.

Inclusion and exclusion criteria

Inclusion criteria were (1) patients undergoing initial surgical resection with a definitive histopathological diagnosis of meningioma (WHO grades I-II), LGGs (WHO grades I-II), or HGGs (WHO grades III-IV); (2) completion of non-contrast head MRI, T1-weighted contrast-enhanced scanning, and APT imaging within one week prior to surgery; (3) availability of complete imaging data with good image quality, free from significant artifacts, suitable for diagnostic evaluation and measurements; (4) availability of complete clinical data and follow-up records.

Exclusion criteria were (1) patients who had received any anti-tumor therapy (e.g., radiotherapy, chemotherapy, targeted therapy, or immunotherapy) prior to surgery; (2) patients with recurrent brain tumors or the presence of other intracranial space-occupying lesions; (3) patients with contraindications to MRI examination; (4) images with severe motion artifacts or susceptibility artifacts that would affect subsequent data analysis; (5) pathological diagnosis of central nervous system tumor types other than those specified above.

Sample size calculation

This study was a retrospective diagnostic analysis, and the sample size was primarily determined by all available cases that met the inclusion criteria during the study period. To evaluate whether the current sample size was sufficient to detect statistically significant differences in APT parameters between groups, we referred to differences in APT signal values between different-grade gliomas and meningiomas reported in previous literature³³. Setting α = 0.05 and a statistical power of 0.80, and using a two-sample mean comparison formula for estimation, we determined that a minimum of 15 samples per group was required. Based on the final inclusion of 105 patients (55 meningiomas, 20 LGGs, 30 HGGs), a post-hoc power analysis showed that the power to detect differences in APTmean among the three groups exceeded 0.95 with the current sample size. This indicates that the study sample size was adequate to support statistical inference.

Image acquisition and post processing

MRI scanning protocol:

All patients were scanned using a magnetic resonance (MR) scanner with a standard 8-channel head coil ³⁴. The scanning sequences and parameters were as follows:

Conventional sequences: Axial T1-weighted imaging (T1WI) was performed using a three-dimensional spoiled gradient-echo sequence (3D T1-SPGR, repetition time/echo time [TR/TE] = 8.5/3.4 ms, flip angle = 12°, slice thickness = 1.2 mm, interslice gap = 0 mm, field of view [FOV] = 24 cm × 24 cm, matrix = 256 × 256). Axial T2-weighted imaging (T2WI) was acquired with a fast spin-echo sequence (TR/TE = 4,500/102 ms, slice thickness = 5 mm, interslice gap = 1.5 mm, FOV = 24 cm × 24 cm, matrix = 384 × 224). Fluid-attenuated inversion recovery (FLAIR) images were obtained with repetition time/echo time/inversion time (TR/TE/TI) = 9,000/120/2,250 ms.

Contrast-enhanced scanning: After intravenous bolus injection of gadopentetate dimeglumine at a standard dose of 0.1 mmol/kg body weight and a flow rate of 2 mL/s, followed by a 20 mL saline flush, axial, sagittal, and coronal contrast-enhanced T1WI images were acquired using the same parameters as the precontrast 3D T1-SPGR sequence.

APT imaging sequence: APT imaging was performed using a two-dimensional gradient-echo pulse sequence with the following parameters: TR/TE = 3,000/3.5 ms, slice thickness = 5 mm, FOV = 24 cm × 24 cm, matrix = 128 × 128, number of excitations = 2, and a total acquisition time of 4 min and 32 s. A continuous-wave saturation pulse with a duration of 2 s and a power (B₁) of 2 µT was applied at frequency offsets ranging from −6 to +6 parts per million (ppm) (in steps of 0.5 ppm), including the water resonance frequency at 0 ppm. Under the main magnetic field (B₀), the saturation pulse was specifically applied at the downfield resonance frequency of +3.5 ppm relative to the water proton resonance (0 ppm), corresponding to the amide proton chemical exchange rate. Prior to APT acquisition, B₀ inhomogeneity was corrected using a water saturation shift referencing (WASSR) approach with saturation parameters identical to the main APT scan but at a reduced B₁ amplitude of 0.5 µT. After scanning, APT-weighted images and magnetization transfer ratio asymmetry (MTRasym) maps at 3.5 ppm were automatically generated by the scanner's built-in software by calculating the asymmetry of the magnetization transfer ratio: MTRasym(3.5 ppm)=[S(−3.5 ppm) − S(+3.5 ppm)] / S(0 ppm) where S(Δω) represents the signal intensity with the saturation pulse applied at the indicated frequency offset Δω relative to the water resonance. The positive MTRasym value at 3.5 ppm primarily reflects the amide proton transfer effect, as the magnetization transfer (MT) and nuclear Overhauser effects are approximately symmetric about the water resonance and thus canceled out by the asymmetry calculation.

Image analysis:

All imaging data were independently analyzed by two neuroradiologists with over 5 years of experience using an image-viewing software workstation. Region of Interest (ROI) delineation was performed according to the following standardized procedures:

Tumor solid region: On the contrast-enhanced T1WI images, an ROI was manually drawn along the enhancing margin of the tumor, carefully avoiding necrotic, cystic, and peritumoral edema areas. The ROI was drawn to encompass the entire solid enhancing component on the slice with the maximum tumor diameter.

T2 Hyperintense region: On the T2WI images, an ROI was manually drawn to encompass the entire hyperintense area, including both the tumor and the surrounding peritumoral edema. The boundary was defined as the visible margin at which the hyperintensity met the normal-appearing brain parenchyma.

APT Abnormal signal region: On the APT-weighted MTRasym maps, using the contralateral normal-appearing white matter as a reference, an ROI was manually drawn to include all areas with signal intensity higher than the mean signal of the reference normal tissue plus 2 standard deviations. This threshold-based approach was applied to the same slice as the maximum tumor diameter. The contralateral reference ROI was placed in the centrum semiovale at a mirror location to the tumor, with an area of approximately 100 mm².

For consistency, each ROI was measured three times on consecutive slices centered at the maximum tumor diameter, and the average value was recorded for analysis. All ROI delineations were performed independently by the two reviewers, and any discrepancies were resolved by consensus. Interobserver agreement was assessed using the intraclass correlation coefficient (ICC), with an ICC > 0.80 indicating excellent agreement.

The following parameters were measured and calculated:

Mean APT Signal Intensity (APT_mean): The average MTRasym value at 3.5 ppm within the ROI of the tumor solid region.

Regional ratio calculation:

RAPT/T2: The ratio of the area of the APT abnormal signal region to the area of the T2WI hyperintense region (RAPT / RT2).

RAPT/E: The ratio of the area of the APT abnormal signal region to the area of the T1WI contrast-enhancing region (RAPT / RE).

Clinical outcome measures

The primary evaluation measures of this study were imaging parameters and their diagnostic efficacy, rather than therapeutic clinical outcomes. Primary Imaging Outcomes: APT_mean values for each group (meningioma, LGGs, HGG), and the RAPT/T2 and RAPT/E ratios for the LGG and HGG groups.

Statistical analysis

Continuous variables were tested for normality using the Shapiro-Wilk test. Normally distributed data are presented as mean ± standard deviation (SD) and compared with one-way analysis of variance (ANOVA) (Bonferroni post-hoc). Non-normal data and ordinal data are presented as M (Q₁, Q₃) and compared using the Kruskal-Wallis H test (Dunn's test with Bonferroni correction). Categorical data are presented as n (%) and compared using the Chi-square or Fisher’s exact test. ROC curves were plotted to evaluate diagnostic performance. All tests were two-sided; P < 0.05 was considered significant.

Results

Baseline and clinical characteristics

Significant differences were observed among the three patient groups in terms of age, Karnofsky Performance Status (KPS) score, initial presenting symptoms, and WHO grade (P < 0.001). Patients in the HGG group were older, had lower KPS scores, and more frequently presented with focal neurological deficits as their initial symptom. In contrast, patients in the LGG group predominantly presented with seizures as their initial symptom. No significant differences were found among the three groups regarding gender, Body Mass Index (BMI), or tumor location (frontal, temporal, parietal, occipital lobes) (P > 0.05) (Table 1).

MRI tumor extent and APT-related parameters

Table 2 presents the comparison results of conventional MRI tumor extent and APT-related parameters among the three patient groups. The degree of contrast enhancement showed significant differences across the groups (P < 0.001), with the meningioma group exhibiting the highest enhancement degree and the LGG group the lowest. No statistically significant differences were observed among the three groups regarding the areas of T2 hyperintensity (RT2), contrast-enhancing region (RE), or APT abnormal signal region (RAPT) (P > 0.05), suggesting comparable overall imaging-defined tumor volumes across the groups.

However, APT-related parameters demonstrated significant intergroup differences (P < 0.001). The mean APT signal intensity (APT_mean) was highest in the HGG group (median 4.60%), significantly exceeding that of both the LGG group (2.95%) and the meningioma group (2.70%). Furthermore, the ratios of the APT abnormal area to the T2 hyperintense area (RAPT/T2) and to the contrast-enhancing area (RAPT/E) were also significantly higher in the HGG group compared to the other two groups.

Pairwise comparisons of APT-related parameters among meningiomas, LGGs, and HGGs

Table 3 presents further pairwise comparisons of the APT-related parameters among the three groups. The results are as follows: Meningioma Group vs. LGG Group: Significant differences were observed in the degree of enhancement, RAPT/T2, and RAPT/E (P < 0.001), whereas no statistically significant difference was found in APTmean (P = 0.241). Meningioma Group vs. HGG Group: Significant differences were observed in APTmean, RAPT/T2, and RAPT/E (P < 0.001), while the degree of enhancement showed no significant difference (P = 0.160). LGG Group vs. HGG Group: Significant differences were observed in the degree of enhancement and APTmean (P<0.001), whereas RAPT/T2 and RAPT/E showed no statistically significant difference (P > 0.05). These findings indicate that APTmean is valuable for distinguishing HGGs from meningiomas and for differentiating glioma grades. Conversely, RAPT/T2 and RAPT/E demonstrate greater discriminatory value in distinguishing LGGs from meningiomas.

APT-related parameters for differentiating meningiomas and LGGs

The results from Dunn’s test indicated statistically significant differences between LGGs and meningiomas in the degree of enhancement (P < 0.001), RAPT/T2 (P < 0.001), and RAPT/E values (P < 0.001). However, the difference in APTmean was not significant (P > 0.05). Therefore, the degree of enhancement, RAPT/T2, and RAPT/E may help differentiate between LGGs and meningiomas. The degree of enhancement was significantly stronger in meningiomas than in LGGs, and the changes observed on APT images due to LGGs were more pronounced than those caused by meningiomas. However, because the APT_mean values for LGGs were close to those for meningiomas, this parameter is not suitable for differential diagnosis in this context.

ROC curve analysis revealed that in differentiating meningiomas from LGGs, the degree of enhancement, RAPT/E ratio, and RAPT/T2 ratio all held significant diagnostic value (all P < 0.001). The degree of enhancement yielded the highest area under the curve (AUC) of 0.926 (95% confidence interval [CI]: 0.842–0.974), demonstrating excellent diagnostic performance. The AUCs for RAPT/E and RAPT/T2 were 0.781 (95% CI: 0.670–0.868) and 0.734 (95% CI: 0.619–0.829), respectively, indicating their good diagnostic performance in distinguishing these two tumor types. It is noteworthy that APTmean showed no significant difference between the two groups (P > 0.05) and is therefore not applicable for differentiating meningiomas from LGGs. These findings are summarized in Table 4 and Figure 2.

APT-related parameters for differentiating meningiomas and HGGs

Dunn’s test results indicated statistically significant differences between HGGs and meningiomas in APTmean (P < 0.001), RAPT/E (P < 0.001), and RAPT/T2 (P < 0.001) values. However, the degree of enhancement did not show statistical significance (P > 0.05). This suggests that APTmean, RAPT/E, and RAPT/T2 may help differentiate HGGs from meningiomas. The APT_mean of HGGs was significantly higher than that of meningiomas, and the extent of abnormality on APT images associated with HGGs was more pronounced than that seen with meningiomas. Since the degree of enhancement for both meningiomas and HGGs is typically moderate or marked, this parameter did not distinguish between these two entities in this cohort.

In differentiating meningiomas from HGGs, APTmean, RAPT/E, and RAPT/T2 all demonstrated excellent diagnostic performance (all P < 0.001). Among them, APTmean achieved the highest AUC of 0.969 (95% CI: 0.906–0.994), indicating a very high discriminatory ability. The AUCs for RAPT/E and RAPT/T2 were 0.836 (95% CI: 0.740–0.908) and 0.800 (95% CI: 0.699–0.879), respectively, further supporting the clinical value of APT imaging in distinguishing these two tumor types. In contrast, the degree of enhancement showed no significant difference between the two groups (P > 0.05) and lacked diagnostic utility for this differentiation. These findings are summarized in Table 5 and Figure 3.

APT-related parameters for differentiating LGGs and HGGs

The results of Dunn's test indicated that the degree of enhancement and APTmean differed significantly between LGGs and HGGs (P < 0.001). However, RAPT/E and RAPT/T2 showed no statistical significance (P > 0.05). This suggests that APTmean and the degree of enhancement are more effective in distinguishing HGGs from LGGs. HGGs typically exhibit a higher degree of enhancement and a higher APTmean compared to LGGs.

In distinguishing between LGGs and HGGs, both the degree of enhancement and APTmean demonstrated significant diagnostic efficacy (both P < 0.001). APTmean achieved an AUC of 0.916 (95% CI: 0.802–0.976), which was superior to that of the degree of enhancement (AUC = 0.823, 95% CI: 0.689–0.917). This indicates that APTmean provided higher diagnostic accuracy for glioma grading in this cohort. In contrast, RAPT/E and RAPT/T2 showed no significant difference between the two groups (P > 0.05) and did not show diagnostic value for grading in this cohort. These findings are summarized in Table 6 and Figure 4.

Data Availability

The original data supporting this study have been submitted as supplementary documents.

Figure 1: Patient selection and imaging analysis workflow. Abbreviations: APT = amide proton transfer; ROI = region of interest; ROC = receiver operating characteristic; T2WI = T2-weighted imaging; CE-T1WI = contrast-enhanced T1-weighted imaging; RAPT = APT abnormal signal area; RAPT/T2 = ratio of APT abnormal area to T2 hyperintense area; RAPT/E = ratio of APT abnormal area to enhanced area. Please click here to view a larger version of this figure.

Figure 2: Differentiation between meningiomas and low-grade gliomas. Receiver operating characteristic curve analysis for differentiating meningiomas and LGGs. Abbreviations: APT = amide proton transfer; LGG = low-grade glioma; RAPT = APT abnormal signal area; RAPT/T2 = ratio of APT abnormal area to T2 hyperintense area; RAPT/E = ratio of APT abnormal area to enhanced area. Please click here to view a larger version of this figure.

Figure 3: Differentiation between meningiomas and high-grade gliomas. Receiver operating characteristic curve analysis to differentiate meningiomas from HGGs. Abbreviations: APT = amide proton transfer; HGG = high-grade glioma; RAPT = APT abnormal signal area; RAPT/T2 = ratio of APT abnormal area to T2 hyperintense area; RAPT/E = ratio of APT abnormal area to enhanced area; APTmean = mean amide proton transfer signal intensity. Please click here to view a larger version of this figure.

Figure 4: Differentiation between low-grade gliomas and high-grade gliomas. Receiver operating characteristic curve analysis for differentiating LGGs and HGGs. Abbreviations: APT = amide proton transfer; LGG = low-grade glioma; HGG = high-grade glioma; APTmean = mean amide proton transfer signal intensity. Please click here to view a larger version of this figure.

Characteristics	Meningioma group (n=55)	LGG group (n=20)	HGG group (n=30)	Statistics	P
Age (years)	54.65±8.21	41.50±10.49	58.17±9.75	F=21.595	<0.001
Sex (n, %)
male	17 (30.91)	10 (50.00)	13 (43.33)	χ²=2.755	0.252
female	38 (69.09)	10 (50.00)	17 (56.67)
BMI (kg/m2)	24.26±3.18	23.59±2.87	24.35±3.25	F=0.412	0.664
Tumor location, n (%)
Frontal lobe	19 (34.55)	8 (40.00)	12 (40.00)	-	0.978
Temporal lobe	13 (23.64)	6 (30.00)	8 (26.67)
Parietal lobe	13 (23.64)	4 (20.00)	6 (20.00)
Occipital lobe	10 (18.18)	2 (10.00)	4 (13.33)
KPS Score, M (Q₁, Q₃)	90.00 (80.00,95.00)	75.00 (70.00,100.00)	70.00 (60.00,90.00)	19.152#	<0.001
Initial Symptoms, n (%)
Headache	23 (41.82)	6 (30.00)	10 (33.33)	-	<0.001
Epilepsy	14 (25.45)	12 (60.00)	4 (13.33)
Focal Neurological Deficit	10 (18.18)	2 (10.00)	16 (53.33)
Other	8 (14.55)	0 (0.00)	0 (0.00)
WHO Grade, n (%)
Grade I	36 (65.45)	16 (80.00)	0 (0.00)	-	<0.001
Grade II	19 (34.55)	4 (20.00)	0 (0.00)
Grade III	0 (0.00)	0 (0.00)	16 (53.33)
Grade IV	0 (0.00)	0 (0.00)	14 (46.67)

Table 1: Patient baseline and clinical characteristics. Note: F = ANOVA; # = Kruskal-Wallis test; χ² = chi-square test; - = Fisher’s exact test. Abbreviations: BMI = body mass index; KPS = Karnofsky Performance Status; WHO = World Health Organization; LGG = low-grade glioma; HGG = high-grade glioma.

Variables	Meningioma group (n=55)	LGG group (n=20)	HGG group (n=30)	Statistics	P
MR information
Enhancement	4.00 (3.00, 4.00)	2.00 (2.00, 3.00)	3.00 (3.00, 4.00)	38.199#	<0.001
Tumor range of MRI (mm2)
RT2	1690.00 (1440.00, 1840.00)	1650.00 (1550.00, 1750.00)	1500.00 (1387.50, 1887.50)	0.898#	0.638
RE	1720.00 (1470.00, 1890.00)	1625.00 (1500.00, 1850.00)	1675.00 (1462.50, 1800.00)	0.729#	0.695
RAPT	1860.00 (1555.00, 2090.00)	1925.00 (1637.50, 2062.50)	2020.00 (1762.50, 2100.00)	3.821#	0.148
APT Parameters
APTmean	2.70 (2.50, 3.10)	2.95 (2.77, 3.23)	4.60 (4.23, 5.25)	54.875#	<0.001
RAPT/T2	102.65 (101.97, 108.15)	114.29 (105.26, 125.00)	116.67 (106.48, 123.81)	23.938#	<0.001
RAPT/E	101.04 (100.78, 105.81)	110.00 (106.06, 113.38)	120.69 (109.60, 123.33)	32.374#	<0.001

Table 2: Comparison of MRI tumor extent and APT-related parameters between glioma and meningioma patients, M (Q₁, Q₃). Note: # = Kruskal-Wallis test. Abbreviations: MRI = magnetic resonance imaging; APT = amide proton transfer; RT2 = T2 hyperintense area; RE = enhanced area; RAPT = APT abnormal signal area; APTmean = mean amide proton transfer signal intensity; RAPT/T2 = ratio of APT abnormal area to T2 hyperintense area; RAPT/E = ratio of APT abnormal area to enhanced area.

Variables	Meningioma group vs. LGG group		Meningioma group vs. HGG group		LGG group vs. HGG group
Enhancement	Z=7.329	P<0.001	Z=1.933	P=0.160	Z=8.102	P<0.001
APTmean	Z=1.749	P=0.241	Z=7.388	P<0.001	Z=4.226	P<0.001
RAPT/T2	Z=3.078	P<0.001	Z=4.564	P<0.001	Z=0.804	P=0.421
RAPT/E	Z=3.303	P<0.001	Z=5.423	P<0.001	Z=1.276	P=0.606

Table 3: Pairwise comparisons of APT-related parameters among meningiomas, low-grade gliomas, and high-grade gliomas. Abbreviations: APT = amide proton transfer; LGG = low-grade glioma; HGG = high-grade glioma; APTmean = mean amide proton transfer signal intensity; RAPT/T2 = ratio of APT abnormal area to T2 hyperintense area; RAPT/E = ratio of APT abnormal area to enhanced area.

Indicators	AUC	SE	95% CI	Z	P
Enhancement	0.926	0.028	0.842-0.974	15.365	<0.001
RAPT/T2	0.734	0.071	0.619-0.829	3.284	<0.001
RAPT/E	0.781	0.053	0.670-0.868	5.287	<0.001

Table 4: The AUC of different indicators for differentiating meningiomas and LGGs. Abbreviations: LGG = low-grade glioma; APT = amide proton transfer; AUC = area under the curve; SE = standard error; CI = confidence interval; RAPT/T2 = ratio of APT abnormal area to T2 hyperintense area; RAPT/E = ratio of APT abnormal area to enhanced area.

Indicators	AUC	SE	95% CI	Z	P
APTmean	0.969	0.017	0.906-0.994	27.488	<0.001
RAPT/T2	0.800	0.051	0.699-0.879	5.865	<0.001
RAPT/E	0.836	0.045	0.740-0.908	7.603	<0.001

Table 5: The AUC of different indicators for differentiating meningiomas and HGGs. Abbreviations: HGG = high-grade glioma; APT = amide proton transfer; AUC = area under the curve; SE = standard error; CI = confidence interval; RAPT/T2 = ratio of APT abnormal area to T2 hyperintense area; RAPT/E = ratio of APT abnormal area to enhanced area.

Indicators	AUC	SE	95% CI	Z	P
Enhancement	0.823	0.055	0.689-0.917	5.851	<0.001
APTmean	0.916	0.053	0.802-0.976	7.909	<0.001

Table 6: The AUC of different indicators for differentiating LGGs and HGGs. Abbreviations: LGG = low-grade glioma; HGG = high-grade glioma; APTmean = mean amide proton transfer signal intensity; AUC = area under the curve; SE = standard error; CI = confidence interval.

Discussion

Accurately distinguishing between gliomas and meningiomas preoperatively and precisely delineating presumed tumor infiltration boundaries on imaging constitute core challenges in neuro-oncology, directly impacting treatment strategies and patients' ultimate prognosis. This study, based on APT, an emerging molecular magnetic resonance technique, systematically evaluated its application value in differentiating gliomas of various grades from meningiomas. Furthermore, by introducing the spatial extent ratio parameter, we thoroughly explored and assessed imaging features related to tumor invasiveness. The results confirm that APT imaging, leveraging its unique biochemical contrast mechanism, has potential as a complement to conventional MRI in differential diagnosis, particularly for distinguishing HGGs, with its signal-intensity parameter APT_mean showing excellent efficacy. More importantly, this study found that quantifying the difference between the spatial extent of the APT abnormal signal and the lesion extent observed on conventional MRI—namely, the RAPT/T2 and RAPT/E ratios—can effectively address the clinical difficulty of differentiating LGGs from meningiomas. This approach also provides a highly promising new imaging method for the non-invasive assessment of the presumed extent of glioma infiltration.

One of the core findings of this study is the potential utility of the mean APT signal intensity (APT_mean) as a biomarker of tumor malignancy, particularly for distinguishing HGG from LGGs and meningiomas. The data from this study show that the APTmean values of HGGs are significantly higher than those of the other two groups. In discriminating between HGGs and meningiomas, the AUC was excellent at 0.969. Furthermore, it exhibited high diagnostic efficacy in distinguishing high-grade from low-grade gliomas, with an AUC value of 0.916. This significant difference in signal is deeply rooted in the tumors' pathophysiological basis. APT technology detects the chemical exchange between amide protons on mobile proteins/polypeptide backbones and water protons within tissues, with its signal intensity directly related to the concentration of these macromolecules and the pH of the tissue microenvironment³⁵. The histological characteristics of HGGs include high cellular density, significant atypia, and active mitotic activity, which are associated with abnormally vigorous protein synthesis and catabolism, leading to a marked expansion of the intracellular pool of freely mobile proteins and polypeptides³⁶. Concurrently, to meet the energy demands of their rapid proliferation, HGGs often exhibit intense aerobic glycolysis, known as the Warburg effect, producing and accumulating substantial amounts of acidic metabolites, such as lactic acid, thereby significantly acidifying the tumor microenvironment. Both the increased protein content and the decreased pH collectively enhance amide proton saturation transfer, ultimately manifesting as a characteristic high signal on APT images³⁷. In contrast, although some subtypes of meningiomas can be highly cellular, the vast majority are benign or low-grade malignancies, with proliferation rates far lower than those of HGGs, and their protein metabolic levels and changes in microenvironmental pH are relatively moderate. LGGs also possess certain proliferative activity, but their degree of malignancy and metabolic intensity cannot be compared to HGGs³⁸. In conventional MRI, both meningiomas and HGGs can present with significant enhancement. However, APTmean can differentiate between them from a metabolic perspective, compensating for the limitations of structural imaging. Moreover, APT imaging can directly reflect microscopic changes in tumor metabolism and may capture areas suggestive of early infiltration and metabolic abnormalities more sensitively than conventional MRI. The elevated APTmean in HGGs is not only associated with the solid tumor region but may also indicate microscopic cellular activity in the peritumoral edema zone, which is radiologically suggestive of infiltration, though this remains to be confirmed by histopathology.

This study revealed the limitations of relying solely on APTmean values to distinguish between LGGs and meningiomas and addressed this issue by using the spatial extent ratio parameters RAPT/T2 and RAPT/E. The results showed no significant difference in APTmean values between LGGs and meningiomas, which may reflect a certain overlap in cellular proliferation activity and protein content between the two. However, when the analytical dimension shifted from signal intensity to spatial distribution, a clear differentiation pattern emerged. The RAPT/T2 and RAPT/E ratios were significantly higher in LGGs than in meningiomas, indicating relatively high diagnostic efficacy, with AUCs of 0.734 and 0.781, respectively. The most fundamental biological characteristic of gliomas, regardless of their grade, is their diffuse, infiltrative growth pattern. Tumor cells can actively detach from the main tumor mass and migrate along anatomical structures such as nerve fiber tracts and perivascular spaces, infiltrating the surrounding, seemingly normal brain parenchyma^39,40. These microscopic foci of cellular dissemination often extend beyond the edema boundaries defined by conventional MRI (T2-weighted or FLAIR sequences) and are far more extensive than the areas of blood-brain barrier disruption visible on contrast-enhanced scans. Due to its high sensitivity to cellular density and metabolic state, APT imaging can detect regions with signal alterations that are radiologically suggestive of tumor cell aggregates in the early stages of infiltration, before they form macroscopic lesions or induce significant edema. This leads to a significantly larger area of abnormal signal on APT images (RAPT) compared to the tumor extent defined by conventional MRI (RT2 or RE). In contrast, meningiomas typically exhibit expansive growth with well-defined margins, featuring a clear boundary between the tumor tissue and the normal brain. Consequently, the extent of their metabolic abnormality (RAPT) largely coincides with the structural or enhancement boundaries seen on conventional MRI, resulting in RAPT/T2 and RAPT/E ratios approaching 100%⁴¹. This quantitative analysis metric, based on differences in growth patterns, effectively differentiates LGGs from radiologically similar meningiomas, offering an innovative diagnostic tool for clinical practice.

Conventional MRI, particularly FLAIR, already provides valuable information on glioma. The T2/FLAIR mismatch sign is a highly specific marker for isocitrate dehydrogenase (IDH)-mutant, 1p/19q-codeleted astrocytoma, aiding preoperative subtyping and grading⁴². FLAIR also defines the peritumoral hyperintense zone, which, though nonspecific, reflects a mixture of edema, infiltration, and gliosis⁴³. When combined with APT imaging, FLAIR helps contextualize the spatial discrepancy between APT abnormalities and conventional T2 hyperintensity, improving radiological inference of infiltration beyond the enhancing core⁴⁴.

Beyond MRI, other modalities offer complementary diagnostic information. Computed tomography (CT) is valuable for detecting intratumoral calcification—more common in meningiomas and oligodendrogliomas, rare in HGGs—which can help when MRI is equivocal⁴⁵. CT perfusion measures cerebral blood volume (CBV): meningiomas typically show markedly elevated CBV, while gliomas show variable patterns, though overlap between hypervascular HGGs and meningiomas limits their standalone utility⁴⁶. ¹¹C-methionine (¹¹C-MET) positron emission tomography (PET) traces amino acid transport and protein synthesis, effectively distinguishing gliomas from meningiomas; importantly, it often reveals abnormal uptake beyond contrast-enhancing or T2/FLAIR regions, analogous to our APT-based spatial ratios⁴⁷. Integrating APT with CT calcification, CT perfusion, and ¹¹C-MET PET could synergistically improve diagnostic accuracy and treatment planning. However, practical considerations such as radiation exposure, cost, availability, and examination time must be carefully weighed when selecting the optimal imaging strategy for individual patients.

Building on the finding that APT imaging can detect tumor-associated signals beyond the boundaries defined by conventional MRI, visualizing and quantifying these signals may be clinically significant. First, in surgical planning, surgeons might, in theory, use APT-defined abnormal signal regions as supplementary information alongside contrast-enhancing lesions and T2 hyperintense areas when selecting surgical targets, which conceptually aligns with "supramarginal resection." However, it is critical to emphasize that this is a radiological hypothesis. The actual presence of tumor cells within the APT-high regions requires validation through image-guided stereotactic biopsy. Without such pathological correlation, APT-defined margins should be viewed as supplementary imaging information rather than confirmed tumor boundaries. This imaging-guided resection strategy theoretically holds the potential to more thoroughly eliminate microscopic residual disease if the APT signal indeed corresponds to tumor infiltration, thereby possibly reducing postoperative recurrence rates and extending patients' progression-free survival and overall survival. Second, in the development of radiotherapy plans, the APT-defined biological tumor volume (BTV) could serve as supplementary information for defining the gross tumor volume/clinical target volume (GTV/CTV) based on traditional anatomical imaging. If validated, this could allow radiation planning to better account for areas with proliferative potential while sparing surrounding normal brain tissue. Third, for stereotactic biopsy of non-enhancing gliomas, APT imaging may help guide biopsy targeting toward areas with the highest signal values. This may increase the likelihood of acquiring diagnostically valuable tissue and reduce pathological undergrading due to sampling error.

The study by Sartoretti et al.⁴⁸ demonstrated that integrating APT imaging with radiomics and machine learning algorithms can further extract its deep-level informational value, achieving high-precision differentiation among HGGs, LGGs, and metastases. Their multilayer perceptron model attained an AUC of 0.836 in distinguishing primary gliomas from metastases. This suggests that APT imaging not only serves as an independent biomarker but also holds promise for its high-dimensional imaging features, when combined with artificial intelligence models in the future, to construct powerful intelligent diagnostic tools capable of automatically segmenting tumors, inferring presumed infiltration boundaries, and predicting molecular subtypes, thereby opening new avenues for the precision diagnosis and treatment of neuro-oncology. Building upon this foundation, the present study further deepens and expands the clinical application boundaries of APT imaging. Our research not only focuses on distinguishing tumor types but also specifically addresses the assessment of imaging features related to the extent of tumor infiltration. By proposing the RAPT/T2 and RAPT/E ratios, we provide imaging evidence that may support non-invasive inference of glioma infiltration boundaries. Furthermore, this study selected meningiomas—tumors often mistaken for gliomas on conventional MRI, especially in atypical cases—as the control group, thereby directly addressing a more challenging clinical differential diagnostic problem. Regarding parameter interpretation, the APTmean and spatial ratio parameters used in this study have clearer pathophysiological correlates, making them easier to understand and adopt in clinical practice than higher-order radiomic features. Therefore, this study not only validates the value of APT in complex differential diagnosis but also facilitates its transformation from a diagnostic tool into a tool that may support surgical and radiotherapy planning.

Successful APT implementation relies on several protocol-critical steps. First, the B₀ correction via WASSR (as described in the Protocol section) is essential; omitting it can introduce spurious MTRasym elevations of 1.5–2.0%. For centers lacking WASSR, alternative field-mapping is less accurate but acceptable. Second, motion artifacts during the ~4.5-minute acquisition are the most common pitfall; we recommend immobilization pads and, if motion is visually evident on MTRasym maps, immediate repetition of the sequence. Third, ROI consistency is operator-dependent—our 2-SD threshold (contralateral normal white matter mean + 2SD) yielded excellent interobserver agreement (ICC > 0.80), and we advise initial training on 10 pilot cases to calibrate threshold selection across different scanner platforms. For clinical adoption, the ~4.5-minute scan time is acceptable in most neuro-oncology protocols; we suggest a phased approach—starting with challenging atypical cases—to build local expertise without disrupting routine workflow. Importantly, these procedural choices directly underpin our results: the standardized saturation parameters (2 s, B₁ = 2 µT) ensured consistent MTRasym measurements across groups, the 2-SD threshold made RAPT/T2 and RAPT/E stable across varying tumor sizes, and consensus reading eliminated individual reader bias. Thus, the methodological rigor detailed in our Protocol is not merely formal but functionally required for the diagnostic performance we observed.

Several limitations of this study should be acknowledged. First, this was a retrospective, single-center study with a relatively limited overall sample size, and notably, the subgroup sizes were imbalanced (meningioma, n = 55; LGG, n = 20; HGG, n = 30). The small number of LGG cases may have reduced statistical power for certain pairwise comparisons and limited the generalizability of our findings. Second, our ROC analyses were derived from the same cohort used for model development, without an independent internal or external validation cohort. This approach is known to potentially overestimate diagnostic performance; thus, our reported AUC values require cautious interpretation and must be confirmed in future prospective, multi-center studies with independent test sets. Third, we did not perform multivariable regression analyses to adjust for potential confounders (e.g., age, tumor size, location, or peritumoral edema volume), as such adjustments would be statistically unstable given the limited sample size per subgroup. Future larger-scale studies should incorporate these covariates to assess the independent contribution of APT parameters. Fourth, our ROC comparisons were exploratory and did not undergo correction for multiple testing; therefore, these results should be viewed as preliminary and hypothesis-generating rather than confirmatory. Fifth, although our imaging findings suggest tumor infiltration beyond conventional MRI boundaries, we lack histopathological confirmation from an image-guided stereotactic biopsy. The actual correspondence between APT signal abnormalities and true tumor cell infiltration remains to be validated. Lastly, APT imaging techniques are still evolving, and variations in field strength, scanner platforms, and post-processing algorithms may affect measurement reproducibility; technical standardization is essential before widespread clinical adoption. Despite these limitations, our study provides a foundation for future investigations integrating APT imaging with radiomics and artificial intelligence to further improve preoperative tumor characterization and surgical planning⁴⁹.

In summary, the multi-parameter APT imaging strategy proposed in this study—integrating APTmean signal intensity with spatial extent ratios (RAPT/T2 and RAPT/E)—provides a supplementary imaging framework for the preoperative differentiation of gliomas from meningiomas and offers imaging-derived information that may serve as a radiological reference for inferring the extent of glioma infiltration. Our findings suggest that this combined approach enhances diagnostic performance beyond conventional MRI, particularly for distinguishing HGGs from meningiomas and characterizing spatial signal abnormalities that may reflect microscopic tumor extension. However, it is critical to emphasize that all interpretations of "infiltration boundaries" in this study are based on imaging surrogates, not on histopathological confirmation. The actual correspondence between APT signal abnormalities and true tumor cell infiltration, as well as their relationship to patient outcomes or survival, requires prospective validation through image-guided stereotactic biopsy and longitudinal follow-up studies. Therefore, while APT imaging holds promise as a non-invasive molecular imaging tool, its current clinical role should be viewed as complementary to conventional MRI rather than as a definitive biomarker for tumor invasion. With further multi-center validation and technical standardization, this approach may contribute to the evolving landscape of precision neuro-oncological imaging and surgical planning.

Materials

List of materials used in this article

Name	Company	Catalog Number	Comments
3.0T MR Scanner	GE Healthcare	Discovery 750W	Used for all conventional and APT imaging sequences.
Contrast Agent (Gadopentetate Dimeglumine)	Bayer Healthcare	Magnevist	Standard dose of 0.1 mmol/kg body weight, administered intravenously at 2 mL/s.
Head Coil	GE Healthcare	Standard 8-channel	Receive-only head coil for signal acquisition.
Image Analysis Software	MicroDicom Ltd.	MicroDicom Viewer (version 3.0)	Used for ROI delineation and parameter measurement by neuroradiologists.
Post-processing Software	GE Healthcare	FuncTool (built-in)	Automatically generates APT-weighted images and MTRasym maps on the scanner.
Statistical Analysis Software	IBM Corp.	SPSS (version 25.0)	Used for all statistical analyses, including ROC curves and group comparisons.