Field strength–dependent sensitivity of chemical exchange saturation transfer: a methodological comparison of 3 Tesla and 7 Tesla in a clinical cohort

Abstract Chemical exchange saturation transfer (CEST) MRI provides insight into tissue metabolism by detecting low-concentration endogenous molecules. While studies at 7 Tesla (7T) have shown enhanced sensitivity and spectral separation, 3 Tesla (3T) remains the clinical standard, and the relative performance of these field strengths in a direct clinical head-to-head comparison remains unclear. This prospective cohort study provides a direct within-subject comparison of multi-pool CEST imaging at 3T and 7T, using age-related tissue changes and glioma molecular subtypes as representative applications of physiological and pathological CEST sensitivity. Forty-three patients (ages 18–76; 18 female) underwent 3T and 7T CEST MRI prior to surgery due to suspected brain tumour; following quality control, 36 datasets were included at 3T and 32 at 7T. CEST amplitudes from amide, amine, aliphatic relayed nuclear Overhauser effect (rNOE) and magnetization transfer pools were quantified in white matter, grey matter, deep grey matter and tumour tissue. A physics-informed conditional autoencoder (PICAE) was applied at 7T to correct B1 inhomogeneity. Age effects were tested using linear regression; tumour subtype differences were tested using the Wilcoxon rank-sum test. The significance level was set to α = 0.05, and the Holm–Bonferroni procedure was applied to correct for multiple testing. At 7T, amide and rNOE showed robust negative correlations with age in grey matter and deep grey matter, supporting the potential of CEST as an ageing biomarker. Age dependence at 3T was weaker, limited to rNOE (grey matter and deep grey matter) and magnetization transfer (white matter and deep grey matter). In contrast, tumour CEST metrics showed no significant age dependence at either field strength. Trends in relative amide contrast were consistent with prior findings but did not reach statistical significance. Sensitivity to tumour molecular subtype was similar across field strengths. Variability analyses showed that conventional 7T processing introduced higher technical variability than 3T, whereas PICAE substantially reduced variability and improved data quality at 7T. In conclusion, 7T CEST MRI demonstrates higher potential as a non-invasive marker of brain ageing, whereas our simplified pipeline did not yield additional information for tumour subtyping at either field strength. These findings underscore both the enhanced sensitivity and the higher technical demands of 7T, and highlight the importance of advanced correction strategies such as PICAE for robust use of single-transmit 7T CEST.