Comparing Multi-Pool Lorentzian Fitting Methods for Guanidinium CEST Quantification in the Human Brain at Clinical 3 T: A Validation Study Using MRS.

PURPOSE: This study aimed to evaluate the performance of our recently proposed double-step multi-pool Lorentzian fitting (DMPLF) method for extracting GuanCEST and other CEST signals in the human brain at clinical 3 T, using MRS as the reference standard. METHODS: A simulation experiment was conducted to compare the performance of DMPLF with single-step multi-pool Lorentzian fitting (MPLF) based on 5,000 Z-spectra generated using the Bloch-McConnell equation. CEST signals extracted using two methods, including APT, GuanCEST, rNOE(-3.5), and MT, were correlated with the proton concentrations used in the simulation. An in vivo study was conducted in 26 healthy volunteers on a clinical 3 T MRI system. GuanCEST signal extracted using two methods was correlated with MRS-measured total creatine (tCr). Regional analyses based on DMPLF-generated CEST maps were performed to evaluate tissue-specific CEST signal differences between gray matter (GM) and white matter (WM). RESULTS: In simulation, CEST signals extracted by DMPLF demonstrated stronger correlations with the corresponding proton concentrations than those extracted by MPLF. In human study, DMPLF generated high-quality CEST maps, and the extracted GuanCEST signal was significantly correlated with MRS-measured tCr (r = 0.66, p < 0.001). Regional tissue analysis showed significantly higher APT and GuanCEST signals in GM than in WM, and higher rNOE(-3.5) and MT signals in WM than in GM. CONCLUSION: The DMPLF robustly extracts multiple CEST signals, particularly the GuanCEST, which shows a strong correlation with MRS-measured tCr. These findings demonstrate its superior performance over the single-step MPLF method and highlight its potential for quantitative CEST analysis in clinical 3 T MRI.