Gliomas account for approximately 80% of primary malignant brain tumors and lead to shorter-term survival than any other type of tumor [1]. Gliomas can be classified into grades I-IV based on histopathological characteristics as reported by the World Health Organization (WHO) classification guideline, in which grade I and grade II are termed low grade and grade III and grade IV are termed high grade [2,3]. Given that the tumor grade is a key factor in guiding the choice of treatment, differentiation between low- and high-grade gliomas is critical [2]. Conventional magnetic resonance imaging (MRI) and dynamic contrast-enhanced MRI (DCE-MRI) can be used to classify gliomas according to morphological features [4] and tumor vascularity [5]. However, conventional MRI alone may not be a reliable classifier [6], and DCE-MRI provides insufficient specificity [7]. Although invasive biopsy provides the gold standard for glioma grading [8], it increases the risks of mortality and major morbidity [9]. Furthermore, the limited number of samples in biopsy may lead to a misinterpretation due to glioma heterogeneity [10]. Therefore, novel MRI techniques possessing endogenous contrast are desired for grading gliomas, complementing conventional MRI.

Chemical exchange saturation transfer (CEST) imaging has shown promising ability to detect diverse metabolites in vivo [11], among which a few have been reported to facilitate tumor diagnosis. Amide proton transfer (APT) imaging, a subtype of CEST imaging, originates from backbone amide protons of endogenous mobile proteins or peptides resonating at 3.5 ppm downfield from bulk water [12], and is a useful biomarker for detecting brain tumors [13]. Glutamate CEST (GluCEST), observed around 3 ppm downfield from bulk water, demonstrated the feasibility of mapping glutamate concentrations in rat brain tumors [14]. The CEST signal at 2 ppm was reported to decrease in brain tumors and was correlated with a reduction in the creatine concentration [15]. Myo-inositol CEST (MICEST) imaging originates from the hydroxyl protons of myo-inositol, which exhibited an altered concentration in brain tumors and showed a myo-inositol concentration dependence between 0.2 and 1.5 ppm downfield from bulk water [16].

The CEST signal is often analyzed by magnetization transfer ratio asymmetry (MTRasym) analysis, in which the reference signal is typically selected from the opposite side of the z-spectrum [17]. However, the MTRasym signal can be interfered with by competing effects, such as direct water saturation (DS) [18], semisolid macromolecular magnetization transfer (MT) [19], aliphatic nuclear Overhauser effect (NOE) [20] and other exchange pools [18]. Several prior studies have attempted to separate the target CEST effect from the other competing effects by using various fit-free CEST metrics, such as MTRnormref [21] and MTRRex [22]. In addition, fit-based methods have also been utilized [23] to obtain cleaner CEST signals.

Given the diversity of the CEST metrics at the different frequency offsets exploited to grade gliomas, there is a need to identify CEST metrics that provide the best differentiation between low- and high-grade gliomas. In this study, we aimed to evaluate the performance of the most frequently-used fit-free CEST metrics (i.e., MTRasym, MTRnormref, and MTRRex) at various frequency offsets and the fit-based CEST metrics, presumably denoting different kinds of metabolites, using semiautomatic regions of interest (ROIs) for newly-diagnosed brain glioma grading in adult humans.