Introduction

Magnetic resonance fingerprinting (MRF) provides the opportunity for simultaneous, multiparametric mapping with drastically undersampled data. Previously we have developed a 2D FISP-based MRF method for preclinical applications at high-fields. It allowed simultaneous T₁ and T₂ mapping in mouse brain with 156-µm in-plane resolution in 2 min acquisition time. However, the method is sensitive to errors in B₁ calibration. Hence, for imaging larger animals such as primates, the accuracy of parameter estimation can be compromised due to B₁ field inhomogeneity. In a recent study, Hamilton et al¹ developed a T₂-prepared MRF method that used small flip angles to achieve B₁ insensitivity. In this study, we aimed to evaluate the accuracy of using T₂-prepared MRF method for 3D high-resolution T₁ and T₂ mapping in monkey brain at ultrahigh field.

Methods

MRF Sequence: Fig. 1 shows the FISP-based, T₂-preprared MRF sequence used in the current study. The sequence consisted of the acquisition of 768 frames partitioned into 16 segments (48 frames/segment). 4 segments were preceded by an inversion pulse and 8 segments were preceded by a T₂-preparation module with the mixing time alternating between 20 and 40 ms. The flip angle in each segment was ramped up sinusoidally from 4° to a maximal value between 6° and 15°. A 600-ms delay was applied between each segment to allow partial signal recovery. Data acquisition used constant TE and TR of 1.7 and 10 ms, respectively. A 4π dephasing per voxel was applied along the slice-selective direction. Total acquisition time of a single fingerprint was ~20 s.
Ex Vivo Imaging Study: The MRF method was evaluated in ex vivo imaging of brain tissue from macaque monkey (Macaca mulatta). 3D MRF data were acquired using a stack-of-spirals trajectory. A variable density spiral readout trajectory was designed to fully sample the inner 20x20 k-space with 48 interleaves and the entire k-space with 96 interleaves. The FOV and matrix size were 90x90x8 mm³ and 256x256x8, respectively, yielding a nominal voxel size of 0.35x0.35x1 mm³. A fully sampled dataset was acquired in ~4.3 hours. T₁ and T₂ maps were obtained from retrospectively undersampled data in the readout plane with an undersampling rate of up to 16. The accuracy of T₁ and T₂ mapping using undersampled data were compared to that using fully sampled data by calculating the normalized root mean square error (NRMSE).

Results

3D T₁ and T₂ Mapping: Results of T₁ and T₂ mapping from fully sampled MRF data are shown in Fig. 2. A large field inhomogeneity was observed at the ventral side of the brain, causing a visible distortion to the T₂ map. With a nominal spatial resolution of 0.35x0.35x1 mm³, both T₁ and T₂ maps showed clear distinction between the white and gray matters. Comparing to in vivo results published in literature² (Table 1), T₁ in the ex vivo brain was significantly lower, ranging from 600 to 700 ms in the gray matter, and from 400 to 500 ms in the white matter. T₂ ranged from 14 to 16 ms in gray matter and from 10 to 12 ms in white matter, which was also significantly shorter than T₂ values in vivo.
Undersampling Capacity: T₁ and T₂ mapping from retrospectively undersampled data are shown in Fig. 3. NRMSE remained within 6% and 15% for T₁ and T₂ mapping, respectively, with an undersampling rate of up to 16 (Fig. 3b). T₁ and T₂ maps generated from 8- and 16-fold undersampled data are shown in Fig. 3c-d. With 16-fold undersampling, the acquisition time can be significantly reduced to 2 min/slice.

Conclusion and Discussion

T₂-prepared MRF enabled 3D T₁ and T₂ mapping of monkey brain at a spatial resolution of 0.35x0.35x1 mm³ with drastically undersampled data. This method provides the opportunity for in vivo parameter mapping at unprecedented spatial resolution with feasible acquisition time.

Acknowledgements

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