Introduction

Fast image acquisition is of great importance for patient comfort and reduction of motion artifacts, particularly for patients with movement disorders such as Parkinson’s disease. A routine clinical scan includes 2D images of various contrasts (i.e. PDw, T₂w, T₁w, and FLAIR). For clinically feasible scan time, images are often restricted to thick slices (~5mm), and small number of slices (~24). Recently MAGiC, a synthetic MR imaging scheme¹ that produces quantitative T₁ and T₂ maps and synthesize the four different contrast images with ~6min scan time was proposed. However, several studies reported degraded FLAIR images due to artifacts such as hyperintense brain surface^2–4. In this study, a 2D quad-contrast sequence that provides the four different contrast images (PDw, T₂w, T₁w, and FLAIR) and two different quantitative maps (T₁- and T₂- maps) of the whole brain (4mm, 32 slices) in 4:14 of scan time is proposed.

Methods

[Quad-contrast sequence] A quad-contrast sequence (Fig. 1) was developed to simultaneously acquire PDw, T₂w, PD-FLAIR, and T₂-FLAIR contrasts. Three methods were used to reduce total acquisition time.
1) Interleaved contrast acquisition: Non-IR-prepped acquisition blocks for PDw and T₂w were located between the inversion RF and IR-prepped acquisition blocks for PD-FLAIR and T₂-FLAIR. An interleaved multi-slice acquisition scheme was applied (Fig. 1b).
2) View sharing: To reduce acquisition block duration, a view-shared turbo spin-echo (TSE) readout was incorporated. The effective echo train length (ETL) was 8 for each contrast, and 4 echoes were shared between the PDw and T₂w; and PD-FLAIR and T₂-FLAIR contrasts (Fig. 1a).
3) Parallel acquisition: GRAPPA acceleration factor 4 was used.
Concatenation of 2 was used to enable larger inversion thickness, in order to minimize inflow artifacts (Fig. 1c). Compared with conventional scans with same scan parameters, the total acquisition was 2.4 times accelerated.
[Quantitative mapping and image synthesis] T₁w images were generated using the following equation ⁵.
$$ T1w=\ \frac{PD_{FLAIR}*PD}{{PD_{FLAIR}}^2+{PD}^2} $$
Quantitative T₁ and T₂ maps were generated by conducting a non-linear least-square fit of the four acquired contrast images to the signal model below.
$$S_{PD}=A\left(1-e^{-\frac{T_{recovery}}{T_1}}\right)$$
$$S_{T_2}=A\left(1-e^{-\frac{T_{recovery}}{T_1}}\right)e^{-ΔTE/T_2}$$
$$S_{PD-FLAIR}=A(1-2e^{-TI/T_1}+e^{-T_{recovery}/T_1}e^{-TI/T_1})\ $$
$$S_{T_2-FLAIR}=A(1-2e^{-TI/T_1}+e^{-T_{recovery}/T_1}e^{-TI/T_1})e^{-ΔTE/T_2}$$
Utilizing the T₁ and T₂ maps, pseudo-contrast images of IR-prepped and T₂w with various TI and TE values were synthesized, analogous to MAGiC. The same exponential model as MAGiC was used for synthesis¹.
Pseudo-FLAIR image with TR/TI/TE=9000/2500/90ms was also synthesized and compared with the conventional- and quad-contrast FLAIR.
[Data acquisition] Quad-contrast images were acquired from a healthy subject (IRB approved; 3T, Trio, Siemens). For comparison, conventional PDw, T₁w, T₂w, and FLAIR contrasts were acquired. The following sequence parameters were common for all acquisitions: FOV=256×256mm², voxel size=1×1mm², slice thickness=4mm, number of slices=32, slice gap=25%. Except for the T₁w contrast image acquisition, GRAPPA acceleration factor=4, and number of ACS lines=30 was used. For the quad-contrast sequence, TR/TI/short TE/long TE=10500/2100/9.1/109ms and total scan time 4:14 was used. Conventional sequences were acquired with the following parameters. PDw: TSE readout, ETL=12, TR/TE=7000/9.4ms, scan time=0:58. T₂w: TSE readout, ETL=12 TR/TE=9000/113ms, scan time=1:14. FLAIR: TSE readout, ETL=12, TR/TE/TI=9000/84/2500ms, scan time=2:26. T₁w: GRE readout, TR/TE=250/2.5ms, flip angle=70°, scan time=2:10. Larger ETL than the quad-contrast sequence (=12) was used to match the effective TE for T₂w while maintaining readout bandwidth.

Results

The total scan time of 6:48 for conventional scans was reduced to 4:14 using the quad-contrast sequence, despite of larger ETL in conventional scans. The four routine contrasts acquired by the proposed sequence and the conventional sequences show comparable tissue contrast (Fig. 2).
The T₁ and T₂ maps, and the synthesized images are presented (Fig. 3, video). The T₁ and T₂ values agree with the literature values (WM: T₁=832±10ms, T2=79.6±0.6ms, GM: T1=1331±13ms, T2=110±2ms)⁶.
The FLAIR images from three different methods (conventional, quad-contrast, and quad-contrast synthetic) are compared (Fig. 4). The quad-contrast FLAIR has relatively lower SNR (=205) than conventional FLAIR (=384). The synthetic FLAIR appears less noisy (SNR not calculated), comparable to conventional FLAIR. Quad-contrast synthetic FLAIR shows hyperintense brain surface, while the natively acquired contrasts do not (Fig. 4, red box).

Conclusion and discussion

In this study, a 2D quad-contrast sequence was developed to reduce scan time and generate PDw, T₂w, T₁w and FLAIR contrast images in a single scan. Utilizing an acceleration factor of 4, the total acquisition time of the four contrasts was reduced to ~4min.
Because the acquired contrasts are intrinsically aligned together, no further registration is necessary. This may be advantageous in the case where alignment between the images is important.
The FLAIR contrast is natively acquired in the proposed sequence. Hence, the quad-contrast FLAIR contrast is not degraded by hyperintense brain surface, which presumably arises from partial volume errors during parameter mapping⁷. In contrast, the synthesized FLAIR displays hyperintense brain surface, similar to FLAIR of MAGiC⁴. The relatively lower SNR of the quad-contrast FLAIR results from the short TI and long TE used. In case SNR of FLAIR is important, synthetic FLAIR can be used. Because T₁- and T₂- maps are fitted using all four contrasts, the noise level of the synthetic FLAIR is not limited to that of native quad-contrast FLAIR.
In conclusion, the proposed quad-contrast sequence reduces the routine clinical MR scan time to 4:14, with whole-brain coverage (32 slices) and quantitative T₁ and T₂ maps.

Acknowledgements

This research was supported by the National Research Foundation of Korea(NRF) funded by the MSIT(NRF-2018R1A4A1025891), and the Brain Research Program through the National Research Foundation of Korea(NRF) funded by the Ministry of Science, ICT & Future Planning (NRF-2019M3C7A1031994).