At 3-T MRI, reliable quantification of CEST contrast is challenged by peak overlap and broad background contamination. Because of this limitation, some previous studies employ four-pool Lorentzian fitting (i.e., amide, direct water saturation, magnetization transfer, and relayed nuclear Overhauser enhancement) for CEST analysis at 3 T, where the multiple CEST contrasts at positive frequency offsets are treated as a single pool. In addition, for fitting-based methods, the fitting accuracy can be readily affected by initial values and boundary conditions. To address these problems, a new CEST analysis approach, named double-step R1rho-based Lorentzian fitting (DROF), with initial value randomization, was developed in this study for extracting five-pool (additional CEST@2 ppm pool) Lorentzian fitting parameters at 3 T. Moreover, a comprehensive evaluation framework including three metrics was proposed to assess the performance of CEST analysis methods. The accuracy and robustness of DROF were evaluated using both simulation and in vivo human data at 3 T, and its performance was compared with existing methods. Results demonstrate that DROF offers improved specificity and stability, suggesting its potential to enhance CEST applicability at 3 T.
Keywords: chemical exchange saturation transfer (CEST); double‐step R1rho‐based Lorentzian fitting (DROF); evaluation metrics; human brain; initial value randomization.
© 2025 The Author(s). NMR in Biomedicine published by John Wiley & Sons Ltd.