B1‐sensitivity analysis of quantitative magnetization transfer imaging
Commonly, off‐resonance qMT imaging uses a magnetization transfer (MT)‐prepared spoiled gradient (SPGR) echo pulse sequence 13. It is a standard SPGR sequence preceded by an off‐resonance radiofrequency (RF) pulse that varies in amplitude and frequency offset between measurements; 10 measurements or more are generally required to fit this Z‐spectrum (normalized MT signal vs. off‐resonance frequencies) 14, and one additional measurement without the MT‐preparation for signal normalization. These qMT techniques also require three additional measurements: B0, B1, and T1. In postprocessing, B0 maps calibrate the off‐resonance frequency of the MT pulse in each voxel. B1 maps are used to scale the SPGR excitation flip angle and MT‐pulse saturation power. A T1 map is necessary to constrain certain fitting parameters of the two‐pool MT fitting model 2. For a given voxel, the measured T1 (T1,meas) is a function of the T1 of the water molecules (T1,f, “f” is for “free pool”) and of the T1 of the macromolecules (T1,r, “r” is for “restricted pool”), and two other parameters (F, ratio of the two pool sizes in the voxel, and kf, the exchange rate constant). The large number of measurements required to sample the Z‐spectrum and additional quantitative maps make qMT a time‐costly technique.
Increasingly, whole‐brain qMT imaging has been achieved via a reduction in qMT measurements 15 and new rapid techniques to measure the required quantitative calibration maps 17. However, integrating new methods into quantitative imaging studies can introduce unintended effects. For example, transitioning from single‐slice T1 mapping techniques (i.e., inversion recovery [IR]) to three‐dimensional [3D] techniques, variable flip angle [ VFA]) also results in transitioning from B1‐insensitive 20 to B1‐sensitive 22 T1 mapping. If VFA is used in the qMT imaging protocol, inaccuracies in B1 will propagate into fitted qMT parameters through two pathways instead of just one (Fig. 1): from errors induced in T1, used to restrict the fitting parameters, and from errors in scaling the MT saturation powers with the B1 maps. The potential effect of B1‐uncorrected qMT on the fitted parameters has been noted in previous work 23; however, these were limited in scope to B1‐insensitive T1 techniques. To our knowledge, no comprehensive characterization of the B1‐sensitivity of qMT (and notably, comparing different T1 mapping methods) has previously been performed.
This work focuses on answering the following three questions: 1) How sensitive is each qMT parameter to B1‐inaccuracies? 2) How does the B1‐sensitivity of qMT parameters differ between protocols using B1‐independent (IR) and B1‐dependent (VFA) T1 mapping methods?; and 3) Which T1 mapping method results in the most robust measure of the pool‐size ratio in the presence of B1‐inaccuracies? To explore these questions, we first focused on simulations under ideal measurement conditions for a single tissue type, and then used this framework to perform a sensitivity analysis of the signal curves.