Synopsis

Here, we present a fast multi-parameter quantitative MRI protocol to estimate R1, R2* and MT saturation in the cervical spinal cord, acquired using only vendor-supplied sequences, in clinically acceptable acquisition times (~11 minutes total). The proposed protocol can be easily adapted to clinical scanners, and requires only stock sequences and freely available post-processing tools, and is faster than a comparable protocol required for PSR mapping in a qMT framework.

Introduction

Recent studies have demonstrated that rapid, high-resolution, quantitative magnetization transfer (qMT) imaging can be performed in the spinal cord (SC), in both healthy subjects¹ and patients², using the single point qMT (sp-qMT) method, to derive estimates of the macromolecular-to-free pool size ratio (PSR). However, the sp-qMT method requires the derivation of empirically determined parameter estimates to constrain the model; therefore, applying the sp-qMT method in new pathologies requires re-validation of the constraints to ensure accuracy across pathological tissue types. Additionally, sp-qMT acquisitions typically utilise custom-built sequences, limiting clinically utility.

Another quantitative MRI approach, multi-parameter mapping (MPM), was successfully applied to map magnetisation transfer saturation (MTsat), longitudinal (R1) and effective transverse (R2*) relaxation parameters of the SC^3,4, but previous studies still required considerable acquisition time, and were performed at a lower resolution than was achieved in the sp-qMT literature^1,2. However, MPM does not require empirical constraints and can provide quantitative indices without the need to reverify the model with each new pathology. In this study, we demonstrate that we can acquire MTsat maps – comparable to sp-qMT-derived PSR maps⁵ – more rapidly than an equivalent sp-qMT acquisition. We show this using only vendor-provided sequences, indicating that MTsat, R1, and R2* maps can be acquired clinically, under 11 minutes.

Methods

Scanning was performed on a 3T Prisma scanner (Siemens, Erlangen, Germany), using a 64-channel head-and-neck coil. Two volunteers (24/26, M) were scanned with the full protocol, MPM-only data was acquired on one patient (42, F) with multiple sclerosis (MS): 12-year disease duration, SC involvement and Expanded Disability Status Scale of 6.5.

MPM data⁶ was acquired using three 3D gradient echo (GRE) scans with predominantly T1 (T1w), proton density (PDw), or magnetization transfer (MTw) contrast. Acquisition parameters, optimised to obtain good quality images of the cervical SC with high in-plane resolution in the shortest possible acquisition time, are detailed in Table I. Two averages were acquired for each contrast to increase SNR. For the MTw scan, a pre-pulse of 9.984ms duration, 1200Hz offset and 500° flip-angle was played out. Phase stabilisation, water selective excitation and gradient distortion correction were applied to improve image quality. The B1+ field was mapped using a 2D TurboFLASH sequence with a preconditioning RF pulse, FOV=256x256x150mm³, 2x2x4mm³ resolution⁷. Total acquisition time for the MPM portion was ~11 minutes. Crucially, these scans were acquired using vendor-supplied sequences, rather than custom sequences previously used.

In order to estimate PSR, two 3D GRE scans were acquired, with, and without MT weighting. Acquisition parameters in Table II. For each scan, six averages were acquired to improve SNR. The duration and offset of the MT pre-pulse were modified to 7.68ms and 2.5kHz, respectively. A field map was collected using a 2D dual-echo gradient echo readout, FOV=192x192x192mm³, 3x3x2mm³ resolution, TR/TE1/TE2/FA=1020ms/10ms/12.46 ms/90˚. Total acquisition time for PSR data was ~19 minutes.

A high-resolution (0.625x0.625x5mm3, 18 slices) MEDIC scan was acquired for registration and segmentation. For all scans, distortions within the SC were minimized by centring a shim volume of 86x45x90mm³ on the cord.

Quantitative R1, R2* and MTsat maps were derived from MPM data using the hMRI toolbox⁸. In brief, maps of R2* were calculated on the signal decay over echoes, using all three acquisitions⁹. Next, the 6/8/8 echoes were averaged, giving SNR comparable to that of the longer qMT scans. R1 was calculated in a variable-flip-angle (VFA) framework, corrected for transmit inhomogeneities using the B1 map, MTsat was calculated by taking saturation into account.

The single-point qMT analysis was performed as described previously^1,2. Briefly, a constrained 2-pool model of the Bloch-McConnell equations was used to generate PSR estimates.

Grey and white matter tissue masks, as well as vertebral labels were derived from the MEDIC scan using the Spinal Cord Toolbox^10,11, and applied to the maps of R1, R2*, MTsat and PSR after co-registration.

Results and discussion

Image quality of mean MPM scans was comparable to qMT scans (Figure 1). MPM maps are shown in Figure 2.

Patient and volunteer MTsat values for vertebral level 3 (no visible MS lesion) are shown in Table III. MPM values show good alignment with previous results. Crucially, the ratio of grey/white matter MTsat values matches that of spqMT-derived PSR.

Conclusion

Acknowledgements

References

1: Smith AK et al: Rapid, High-Resolution Quantitative Magnetization Transfer MRI of the Human Spinal Cord, Neuroimage, 2014

2: Smith AK et al: Evaluating single-point quantitative magnetization transfer in the cervical spinal cord: Application to multiple sclerosis

3: Grabher P et al: Tracking sensory system atrophy and outcome prediction in spinal cord injury, Annals of Neurology, 2015

4: Freund P et al: MRI investigation of the sensorimotor cortex and the corticospinal tract after acute spinal cord injury: a prospective longitudinal study

5: Helms G et al: High-resolution maps of magnetization transfer with inherent correction for RF inhomogeneity and T1 relaxation obtained from 3D FLASH MRI, Magn Reson Med, 2008

6: Weisskopf N et al: Quantitative multi-parameter mapping of R1, PD*, MT, and R2* at 3T: a multi-center validation, Front Neurosci, 2013

7: Chung S et al: Rapid B1+ mapping using a preconditioning RF pulse with TurboFLASH readout, Magn Reson Med., 2010

8: Balteau E et al: hMRI - A toolbox for using quantitative MRI in neuroscience and clinicalresearch, WIAS, 2018 (pre-print)

9: Weisskopf N et al: Estimating the apparent transverse relaxation time (R2*) from images withdifferent contrasts (ESTATICS) reduces motion artifacts, Front Neurosci, 2014

10: De Leener B et al: SCT: Spinal Cord Toolbox, an open-source software for processing spinal cord MRI data, NeuroImage, 2017

11: Perone CS et al: Spinal cord gray matter segmentation using deep dilated convolutions, Arxiv, 2017