Specific APT and NOE Imaging Using DSP-CEST in Humans at 3 T.

Abstract

Listen

Translate article

Specific chemical exchange saturation transfer (CEST) imaging, especially at low-field clinical 3-T scanners where multiple pools overlap, is challenging. Recently, a double saturation power (DSP)-CEST method has been developed to improve the specificity in quantifying two major CEST variations: amide proton transfer (APT) and nuclear Overhauser enhancement (NOE). This method avoids the mutual interference between APT and NOE, unlike the conventional asymmetry analysis method. It is also model-free, making it more robust compared to various Lorentzian fitting methods. Initially, this method was developed for continuous-wave saturation and tested in animals on preclinical MRI. This paper aims to further develop the method for pulsed-CEST saturation and apply it to human imaging at 3 T. Simulations and phantom experiments were first conducted to validate its specificity. Subsequently, invivo experiments were performed on six healthy human brains to demonstrate its applications. Two CEST quantification metrics, including the magnetization transfer ratio (MTR) and apparent exchange-dependent relaxation (AREX), were evaluated. The DSP-CEST successfully eliminated all confounding components and specifically quantified the APT and NOE effects. It demonstrated significantly lower MTR-quantified APT effects and higher AREX-quantified APT effects in white matter (WM) than in gray matter (GM). Moreover, it revealed no significant difference in the MTR-quantified NOE effect between WM and GM, but a significantly higher AREX-quantified NOE effect in WM than in GM. These results, differing from those obtained using the asymmetry analysis, confirmed their different origins. In contrast, they align with those obtained using a Lorentzian difference (LD) analysis, suggesting both methods have improved specificity. Yet, without the need to assume the fitting models and set fitting parameters, the DSP-CEST method proves to be more robust than the LD method.