Introduction

3D magnetic resonance fingerprinting (MRF) allows whole-brain T₁ and T₂ mapping at 0.8 to 1 mm resolution in human subjects.^1,2 In combination with parallel imaging, an acceleration factor of 144 can be achieved. Translation of 3D MRF to preclinical studies has been challenging because of the requirement for higher spatial resolution and the lack of parallel imaging capability. In this study, we aimed to develop a 3D MRF method for high-resolution, whole-brain T₁ and T₂ mapping in monkeys. A conformal head coil was used for improved SNR, which enabled 0.35x0.35x1 mm³ resolution with 24-fold acceleration.

Methods

MRF Sequence: The MRF sequence consisted of FISP acquisitions of 768 frames partitioned into 16 segments (48 frames/segment) (Fig. 1). 4 segments were preceded by an adiabatic inversion pulse with an inversion time (TI) of 21, 56, 400, and 150 ms, respectively. 8 segments were preceded by a T₂-preparation module with the mixing time (TM) alternated between 45 and 60 ms.^3,4 Data acquisition used flip angles ramping up sinusoidally from 6° to a maximal value between 8° and 17° in each segment. Constant TE/TR of 1.7/10 ms was used. Acquisition time of a single fingerprint was ~21 s.
3D MRF data were acquired using a stack-of-spirals trajectory. The spiral trajectory sampled the entire k-space in 96 interleaves with an FOV of 90x90 mm² and a matrix size of 256x256. A dephasing gradient was used to induce a 4π dephasing per voxel along the slice-selective direction.
B₁-Corrected Dictionary Matching: Dictionaries with a B₁ factor ranging from 0.2 to 1.5 were simulated. When matching an acquired fingerprint to the dictionary, the B₁ factor from a separately acquired B₁ map was used to select the dictionary for matching. Normalized inner product between the acquired fingerprint and the simulated dictionary was used to find the best match to derive the T₁ and T₂ values.
Phantom Validation: All MRI studies were performed on a Bruker 9.4-T scanner. The MRF method was evaluated on a 7-compartment phantom with MnCl₂ concentration varied from 40 to 400 μM. The accuracy of T₁ and T₂ mapping was first evaluated using an 80-mm birdcage coil without the cofounding B₁ inhomogeneity. The efficacy of B₁ correction was then assessed using a custom-built 100-mm conformal head coil. A fully sampled dataset with an FOV of 90x90x8 mm³ and a matrix size of 256x256x8 was acquired in ~4.3 hours. B₁ maps were acquired using the double-angle method with 45° and 90° FA, respectively.⁵ T₁ and T₂ maps acquired using the inversion recovery spin-echo (IR-SE) and spin-echo (SE) method, respectively, were used as the gold-standard for comparison.
Ex Vivo and In Vivo Studies: MRI studies were performed ex vivo on an excised monkey brain, and in vivo on four female macaques with 4 and 6 years of age (N = 2/age group). 3D MRF data were acquired using the conformal coil with an FOV of 90x90x80 mm³ and a matrix size of 256x256x80, leading to a spatial resolution of 0.35x0.35x1 mm³. Total acquisition was ~1.4 hours with an acceleration factor of 24 (12 in-plane and 2 through-plane).¹ B₁ maps were acquired in 0.4 hours. IR-SE and SE scans were also acquired from ex vivo brain.

Results

Phantom Results: Fig. 2 shows the comparison of T₁ and T₂ mapping using the conformal and birdcage coil. While T₁ mapping showed insensitivity to B₁ inhomogeneity, T₂ overestimation (>5%) occurred when there was a >15% deviation in B₁ from its nominal value. Applying B₁ correction in dictionary matching can effectively reduce the T₂ overestimation.
Ex Vivo Results: T₁ maps by MRF and IR-SE showed less than 5% difference in all selected ROIs (Fig. 3). Consistent with the phantom results, T₂ overestimation was evident in regions with significant B₁ inhomogeneity (cerebellum and right cortex) if no B₁ correction was applied. T₂ maps by MRF after B₁ correction showed good agreement with SE in most structures of gray matter. However, there was ~20% underestimation in regions with short T₂ such as in the white matter and cerebellar nucleus, which might be related to the long mixing time used in T₂ preparation modules.
In Vivo Results: T₁ and T₂ of selected ROIs in ex vivo and in vivo monkey brains are summarized in Table 1. Both T₁ and T₂ measured in vivo are longer than that measured ex vivo. In vivo T₁ was in the range of 1200 to 2000 ms in the brain and ~3500 ms in the CSF. T₂ varied between 20 and 40 ms in the brain and was ~215 ms in the CSF. Comparing T₁ and T₂ values in the two age groups, there was ~20% T₂ reduction in the area of globus pallidus in 6-year-old monkeys (Fig. 4).

Discussion and Conclusion

The current study presents the first in vivo whole-brain T₁ and T₂ mapping of monkey at high spatial resolution. The high SNR provided by the conformal coil, in combination with B₁ correction, enabled accurate T₁ and T₂ mapping with 24-fold acceleration. The reduced T₂ in globus pallidus at older age is in agreement with a previous human study, possibly related to increased iron deposition.⁶ Further acceleration can be achieved by sequence optimization.

Acknowledgements

This work was supported by a grant from the National Institute of Health (R01 EB23704).