Introduction

Myelin-Water Fraction (MWF) mapping has shown great potential in characterizing brain’s myelination and demyelination. Conventional MWF-mapping utilizes a multi-echo spin-echo(1) or gradient-echo sequence(2) and multi-compartment fitting of the exponential decay to extract the shorter relaxation time of myelin-water. However, such fitting is ill-conditioned and prone to noise. To improve MWF-mapping, the ViSTa technique(3,4) was proposed for direct visualization of myelin-water. Here, a specifically configured double inversion-recovery sequence is employed, where the long T₁-component is suppressed and those signals from short T₁-components of myelin-water are preserved. The key challenges of ViSTa are the decreased SNR from the signal suppression and the long acquisition time. Recently, MWF-mapping has also been performed using a modified MR Fingerprinting sequence(5,6), specifically designed to provide good signal separability between myelin-water and other tissue types. Nonetheless, multicompartment fitting is still required to extract MWF, which can be challenging, necessitating the use of predefined compartments with fixed T₁ and T₂ values(7).

In this work, we incorporate ViSTa into MR fingerprinting (ViSTa-MRF), to improve the speed and accuracy of MWF-mapping. With the ViSTa inversion-preparation, the first time-point of ViSTa-MRF will contain direct myelin-water signal. By utilizing spatiotemporal subspace reconstruction(8,9), the spatio-temporal encoding information across the whole ViSTa-MRF time-series is employed to reconstruct this direct myelin-water data at high acceleration, providing MWF-mapping without need for multicompartment modeling. Using ViSTa-MRF along with an efficient 3D-spiral-projection acquisition (10), we demonstrate high-quality whole-brain 1mm³ MWF and T₁/T₂/PD maps in ~10 minutes.

Methods

Pulse sequence: Figure1(A) shows the diagram of the proposed ViSTa-MRF sequence, where each acquisition group consists of eight ViSTa blocks and one MRF block. In each ViSTa block, a double inversion-recovery is performed, with the first subsequent signal time-point labeled the “ViSTa signal”. Through extended-phase-graph (EPG) simulation(11), Fig.1(B) shows that the myelin-water signal with short-T₁ is preserved in the ViSTa signal, while the white-matter (WM), gray-matter (GM) and CSF are suppressed, which enables direct myelin-water imaging. Figure1(C) shows the ViSTa-MRF signal curves with good signal separability between the different tissue types. To increase the emphasis on the encoding of the short-T₁ signal, the ViSTa-MRF sequence acquires forty FISP-readouts with 30° excitations after each ViSTa block and repeat the block eight times with different spatial-encodings, as reflected on the signal curves in Fig1.(C). After the ViSTa blocks, 500-time-point FISP-MRF data are acquired. Between the acquisition blocks, a 90° saturation pulse with a waiting time of 380ms is used to saturate flow effects and achieve steady-state longitudinal magnetization. A total of 820 time points is acquired in each acquisition group across 19.4s.

Acquisitions: 3D-spiral-projection trajectory was used for ViSTa-MRF acquisition at 1-mm whole-brain on a 3T Siemens Prisma scanner with a 64 channel head-neck coil: FOV:220×220×220mm³, TR/TE=12/2.5ms with a 7.0ms spiral readout. 32 acquisition groups were acquired, where the in-plane 16x-undersampled spiral interleaves in each group were designed to rotate around kx, ky, and kz axes by tiny-golden-angle to maximize the incoherence between each time point (Fig.2(A)). This resulted in a scan time of 19.4s×32 = 10.3 minutes, where the acceleration factor of each time point is 16×(220/2)×π/32=172.8. With 8 ViSTa blocks, the acceleration factor of the ViSTa signal is 172.8/8=21.6.

Post-processing: Subspace reconstruction was performed, where ViSTa-MRF dictionary was generated and the first eight principal components were selected as the temporal basis Φ shown in Fig2(A). The desired ViSTa-MRF time-series x is expressed as x=Φc, where c are the temporal coefficient maps. Figure2(B) shows the flowchart of the subspace reconstruction, which could be described as:
$$ min_{c}\bf || MFSΦc|| +\lambda_{1}||c||_{2}+\lambda_{1}||c||_{*} , $$
where S are coil sensitivities, F are the non-uniform FFT operators, M are the undersampling patterns, λ₁ and λ₂ are the regularization parameters. As shown in Fig.2(C), the reconstructed c are used to generate the ViSTa-MRF time-series. Template matching is then applied to extract T₁/T₂/PD maps, while the quantitative MWF map is derived using the first time-point ViSTa image I_(ViSTa) and the PD map I_(PD):
$$ MWF =\frac{I_{(ViSTa)}}{I_{(PD)}\cdot S_{(myelin-water)}} , $$
where S_{(myelin-water)} is the EPG-simulated signal intensity in ViSTa image using nominal T₁(120ms) and T₂(20ms) values of myelin-water. All post-processings were performed using MATLAB and BART(12).

Results

Figure3 shows the ViSTa images reconstructed by parallel imaging and subspace reconstruction at R=21.6. Compared to a direct SENSE reconstruction of the ViSTa data, the joint subspace reconstruction of ViSTa and MRF time-series significantly reduces aliasing artifacts on the ViSTa image, demonstrating the benefit in leveraging the joint spatiotemporal encoding information.

Figure4 shows whole-brain 1mm³ T₁/T₂/PD/ViSTa and MWF maps in three orthogonal views, where high-quality results are observed from this fast acquisition.

Figure5 shows the MWF maps from two representative slices, where the MWF values from ViSTa-MRF across four WM regions are consistent with the literature results(3). Low-frequency signal bias can be observed in our MWF results, likely from B₁+ inhomogeneity and the large FAs used in the MRF-ViSTa sequence (ranged from 10°-82°), which we aim to correct in future work as per(13).

Discussion and conclusion

In this work, we developed a 3D ViSTA-MRF sequence together with subspace reconstruction to achieve rapid high-isotropic-resolution MWF and T₁/T₂/PD mapping in a single scan. In comparison to conventional myelin-water imaging, the proposed method can provide improved-SNR, faster, and high-quality parametric mapping.

Acknowledgements

This work was supported in part by NIH research grants: R01-EB020613, R01-EB019437, R01-MH116173, P41EB030006, and U01-EB025162.