Introduction

Neuromelanin (NM)-sensitive MRI is a promising method for Parkinson’s disease^1–3. Recently, MT-prepared GRE was shown to provide good NM contrast^4–6. While an MT pulse is available in all vendors, the parameters that govern MT efficiency (e.g. FA, offset-frequency) are different and are not flexible. In this study, we propose a new NM-sensitive imaging protocol, sandwich-fsNM, that utilizes multiple flow saturation (FS) RF pulses for MT weighting. The FS pulses, which are flexible (number of pulses and offset-frequency), are located both superior and inferior to imaging volume, avoiding asymmetric MT effects. We demonstrate the effectiveness of our method by simulation and experiments.

Methods

[MT effects of FS pulse] MT effects of an FS pulse are depicted in Figure 1. The FS pulse is applied with a slice selection gradient; therefore, the offset-frequency is a function of the position of a slice of interest (dS) from the top of the imaging volume:
$$Offset-frequency(dS)\,=\,(0.5S_{flowsat}\,+D_{flowsat}\,+dS)\times\gamma G_{slice}$$
where S_flowsat is FS thickness and D_flowsat is FS gap. This means users can determine the offset-frequency which is an advantage over most product MT pulses.

[Two-pool Bloch simulation] The signal intensity of a spin at the position dS can be simulated by two-pool Bloch simulation⁷ (Figure 2). A super-Lorentzian absorption lineshape was chosen, and the simulation parameters for crus cerebri (CC) and substantia nigra (SN) were adapted from a previous research⁵. M₀^f of CC and SN were set to be 0.86 and 1, respectively.

[Sequence parameters] Images were acquired in three healthy subjects (IRB approved; 3T, Trio, Siemens). The sequence parameters for fsNM were fixed as follows unless mentioned otherwise: FOV=230×230×40mm³, voxel size=0.5×0.5×2.5mm³, TR/TE=60/4.77ms, FA=20°, and readout bandwidth=170Hz/pixel, S_flowsat=50mm, D_flowsat=0mm, number of FS (N_flowsat)=6. The FS pulse had duration=3.840ms and RF bandwidth=2176Hz.

[MRI experiments] Four experiments were conducted to determine the sequence parameters that provide the best NM contrast. For Experiment 1 to 3, the two-pool Bloch simulation was conducted using the specified sequence parameters.
Experiment 1: single-sided FS vs. sandwiched FS
To evaluate the symmetricity of the double-sided (sandwiched) FS (e.g., when N_flowsat=2k, k FS regions located inferior to and others superior to the imaging volume), fsNM was acquired with six single-sided (inferior to imaging volume) FS and six sandwiched FS. For the case of the sandwiched-FS, the superior and inferior saturation pulses (three each) were alternately applied.
Experiment 2: FS gap
To assess the effects of the gap between the FS region and imaging volume, the sandwich-fsNM acquisition was repeated with D_flowsat=0, 20, and 40mm. N_flowsat was fixed to 2.
Experiment 3: number of FS
To assess the effects of N_flowsat in the images, the sandwich-fsNM acquisition was repeated with N_flowsat=2, 4, and 6.
Experiment 4: comparison with conventional NM
For comparison, two conventional NM-sensitive images were acquired in a single subject: 3D GRE and 2D TSE images both with MT pulse preparation (offset-frequency=1200Hz, FA=500°, RF bandwidth=230Hz) and T1 weighting (MT-GRE and MT-TSE). The MT-GRE sequence parameters were matched to the fsNM parameters, and MT-TSE images were acquired with FOV=230×230mm², slice thickness 2.5mm, voxel size=0.5×0.5mm², TR/TE=910/14ms, and readout bandwidth=120Hz/pixel.

[Contrast ratio] Contrast ratio (CR) between SN and CC was calculated for both simulation and experimental data as described in a previous study⁸:
$$CR(\%)=\frac{(I_{SN}-I_{CC})}{I_{CC}}\times100$$
where I_SN and I_CC are the simulated signals in case of simulated data, or the ROI-averaged signal intensities in the case of experiment.

Results

The simulation results demonstrate more symmetric MT effects in the sandwiched FS than in the single-sided FS (Figure 3a,b). The experimental results confirm this finding (Figure 3b): In the superior slice, the single-sided fsNM image shows smaller CR between SN and CC (22.4%) than that of the sandwich-fsNM image (30.7%) whereas the intermediate slice shows similar contrast in both acquisitions.
The effects of D_flowsat and N_flowsat are summarized in Figure 4. The simulation and experimental results reveal increased CR as D_flowsat decreases (Figure 4a) and as N_flowsat increases (Figure 4b). When experimental data are further analyzed, the CNR increases and the SNR decreases, as D_flowsat decreases and as N_flowsat increases (Tables in Figure 4c, d). Both SN and LC are most clearly visualized when D_flowsat=0mm and N_flowsat=6. The effect size of changing D_flowsat (3% difference from D_flowsat=40 to 0mm) is smaller than changing N_flowsat (15.5% difference from 2 FS pulses to 6).
When compared with the MT-GRE and MT-TSE images, our sandwich-fsNM images display better contrast both visually and by metrics (Figure 5). When averaged across slices, CR (or CNR) between SN and CC of sandwich-fsNM is 1.66 (or 1.34) times higher than that of MT-GRE, and 1.14 (or 1.79) times higher than that of MT-TSE (Figure 5b). The SNR of sandwich-fsNM is slightly lower than that of MT-GRE, but larger than that of MT-TSE (Figure 5b).

Discussion and Conclusion

Sandwich-fsNM provides higher contrast than the conventional methods, allowing clear delineation of SN and LC. Conventional MT pulses (e.g., offset-frequency=1200Hz, FA=500°, saturation-BW=230Hz) benefit from high FAs whereas FS pulses (e.g., offset-frequency=2629Hz (flexible), FA=90°$$$\times$$$6 or $$$\times$$$4, saturation-BW=2176Hz) have flexible offset-frequency and number of pulses, and wider saturation bandwidth, providing higher MT effects than the conventional MT pulse. This flexibility can be beneficial in cross-vendor studies.

Acknowledgements

This work was supported by Creative-Pioneering Researchers Program through Seoul National University (SNU), the National Research Foundation of Korea (NRF) grant funded by the Korea government (MSIT) (No. NRF-2018R1A2B3008445), and partly by Heuron.