Introduction

Quantitative magnetization transfer (qMT) imaging has shown potential for the prognosis and diagnosis of different neurological abnormalities, such as multiple sclerosis (1). Estimation of qMT tissue parameters from a two-pool model system, however, is currently limited in the clinical setting by long data acquisition times (2). Recently, for rapid whole-brain magnetization transfer ratio (MTR) imaging, the use of a prototype multi-slice MT-weighted spoiled gradient echo (SPGR) sequence was suggested, offering whole brain MTR imaging with intrinsic B₁-correction within less than one minute (3). The method takes advantage of a rapid interleaved spiral readout offering whole-brain coverage for the acquisition of a single MT-weighted volume within 20 seconds. The same underlying spiral SPGR prototype was also used for rapid B₁-corrected whole-brain T₁ mapping, again in less than one minute (4). In this work, we propose to fuse both concepts for rapid in-vivo qMT parameter estimation from a two-pool model system. We will show that the spiral prototype offers whole-brain B₁-corrected quantitative estimation of all qMT tissue parameters within less than 5 minutes.

Methods

MT-weighted whole-brain volumes are acquired using a rapid multi-slice SPGR prototype sequence, as described in (3,4). In this work, a single high-resolution MT-weighted volume is reconstructed from the acquisition of 50 interleaved slices with a TR of 1650 ms. For each acquisition, the sequence kernel (effective TR of 33 ms) consisted of a MT-preparation module (with gaussian-shaped saturation flip angle β, offset frequency Δ, and duration of 17.92 ms) followed by the slice excitation (3mm thickness, sinc-shaped RF pulse of 0.6 ms duration, and flip angle of α=25°) using an interleaved spiral readout (14.2ms duration). For each slice, 20 interleaved spiral trajectories were acquired and reconstructed to a 1.3 x 1.3 mm² in-plane resolution using a spiral version of the iterative self-consistent parallel imaging reconstruction method (SPIRiT) in combination with parallel imaging of factor two (5). From the multi-slice acquisition, the repetitive excitation, and the interleaved spiral readout, the spin magnetization of a single slice effectively preceded a pulse train of 50 MT-saturation pulses. The overall acquisition time for a single MT-weighted whole-brain volume was 20 seconds (including a dummy preparation period of 4s without readout but with MT-preparation and excitation pulses to put the two-pool system in the steady-state). For an overview, see Figure 1.
In addition, a B₁-field and an observable T₁-map (T_1,obs) were acquired based on the same spiral prototype sequence, described in (3). Essentially, the same resolution was used as for the MT-weighted scans described before. The overall acquisition time for the T₁-B₁ mapping was 53 seconds.
Two-pool model parameter estimation was performed from a set of 12 MT-weighted volumes in combination with the T_1,obs and B₁ map as described in (6). To this end, a super-Lorentzian line-shape was used and, since the model is not sensitive to the longitudinal relaxation rate of the bound pool fraction (R_1r), it was further assumed that R_1r = R_1f = 1/T_1,obs. The set of MT-weighted volumes were acquired with Δ = 1kHz, β = 150°,250°,350°,450°,550°,650°,750°,850° and Δ = 6kHz, β = 250°,450°,650°,850°. The MT flip angle was modulated by the corresponding B1-field prior to the estimation of remaining two-pool model parameters, that is the bound proton fraction (F), the exchange rate (k_f), and the transverse relaxation time of the bound pool protons (T_2r).
Three volunteers were scanned at 3T (Magnetom Prisma, Siemens Healthcare, Erlangen, Germany) using a 20-channel receive head coil. Measurements were approved by our local ethics committee. Image postprocessing was conducted by MATLAB R2019a (The MathWorks, Inc., Natick, MA). The standard software package FSL (FMRIB Software Library v6.0, Oxford, UK) was used for co-registration of images. Whole-brain imaging of 12 MT-weighted images together with T₁-B₁ mapping was completed within 4 minutes and 53 seconds.

Results

Figure 2 shows all the retrieved quantitative parameter estimation for three exemplary axial slices, that is F, k_f, T_2r, T_1,obs, and B₁. Coronal and sagittal views of the results are presented in Figure 3. For a selected region-of-interest (ROI) in white and gray matter (cf. Figure 2), parameter estimations are given in Table 1. Overall, the retrieved two-pool model parameters are in good agreement with the literature (7).

Discussion

Recently, a spiral prototype sequence was proposed for rapid whole-brain B₁-corrected T₁ and MTR quantification in less than one minute. In this work, we developed this approach further to offer rapid whole-brain qMT imaging in clinically relevant acquisition times. The T₁-B₁ mapping method yields a simultaneous B₁-map that was used not only to remove the bias in the T₁ estimate but also to improve the overall accuracy of the two-pool model parameter estimation. In this work, we targeted robust whole-brain qMT imaging in clinically relevant acquisition times, within 5 minutes. To this end, sampling of the MT-response was confined to 12 acquisitions with two different frequencies and a variety of different MT flip angles.

Conclusion

In conclusion, we have shown that whole-brain B₁-corrected qMT imaging is feasible within less than 5 minutes with clinically relevant resolutions at 3T using an interleaved spiral prototype sequence. The proposed approach thus offers excellent prospects for wide clinical translation and use.

Acknowledgements

This work was supported by the Swiss National Science Foundation (SNF grant No. 325230_182008).