Functional MRI of the spinal cord is challenging due to its small size and deep anatomical location. Increasing field strength enhances BOLD signal and improves SNR, but B0 distortions produced by the lungs and vertebral column are amplified, presenting additional challenges in protocol optimization. Barry et al. have successfully performed resting-state fMRI at 7T; here, we present observations of robust, well-localized motor task activation in the human cervical spinal cord at 7T. We assessed single-shot and multi-shot EPI at two different resolutions. Multi-shot EPI achieved finer resolution and less spatial distortion in this preliminary 7T spinal cord task fMRI study.
Data were acquired in the cervical spinal cord (C4-C7 vertebral levels) of a healthy volunteer using a 7T whole-body MRI system (Magnetom, Siemens, Erlangen, Germany) and a volume-transmit, 16-channel receive cervical spine RF coil (Quality Electrodynamics, Mayfield Village, OH, USA). Three gradient-echo EPI protocols (single-shot 0.75mm resolution, and 4-shot 0.75mm/0.60mm resolution) were assessed; parameters are tabulated in Figure 1. Pulse-oximeter and respiratory traces were simultaneously acquired. Multi-shot reconstruction included a navigator-based per-shot frequency offset demodulation to mitigate respiratory field variations.
For each 6min 30s experiment, the subject performed 12 blocks of 20s rest and 10s task (tapping the thumb and first three fingers of the left hand at 3Hz), with a terminal 20s rest period. This task was expected to produce sensory activation in the ipsilateral dorsal horn at the neurological C6-C7 (vertebral C5-C6) level, and motor activation in the ipsilateral ventral horn at the neurological C7-C8 (vertebral C6-C7) level.
Four-dimensional image series were motion-corrected slicewise (x- and y-translation) using FSL FLIRT [7], and temporal SNR (tSNR) was calculated voxelwise as the temporal mean divided by temporal standard deviation. GLM analysis was performed within a mask of the spinal cord in FSL FEAT [8], incorporating a 37-term physiological noise model (8 cardiac, 8 respiratory, 16 interaction, heart rate, respiratory volume per time, CSF signal, and 2 motion correction terms) [9].
Temporal mean images with overlaid activation maps for all three protocols are shown in Figure 2. Motor activity is observed in the ipsilateral ventral horn at vertebral level C6 in all three datasets (Fig. 2a,c,e, green arrows), while sensory activity is visible in the ipsilateral dorsal horn at vertebral level C5 in the single-shot dataset only (Fig. 2a, cyan arrow).
Plots of BOLD signal and model fits are shown in Figure 3 for the voxel with the greatest z-score within each dataset (Fig. 2a,c,e, green arrows). The highest z-score is observed in the single-shot 0.75mm dataset, while BOLD percent signal change (%SC) increases with effective image resolution (Fig. 4).
Focally high temporal SNR is observed in the center of the cord in the single-shot images (Fig. 2b), but tSNR averaged within the entire spinal cord is greatest in the 4-shot 0.75mm images (Fig. 4). The tSNR drops markedly towards lower slices in all three protocols (Fig. 5).
The single-shot protocol has the highest steady-state NMR signal owing to the long TR, but is also more vulnerable to field inhomogeneities due to the long readout and TE. This leads to geometric distortion and signal loss, resulting in low tSNR especially at the periphery of the spinal cord and in lower slices. The effective resolution is also diminished by partial Fourier encoding and T2* blurring. Hence, relative to the single-shot image, the 4-shot images demonstrate greater resolution, greater BOLD %SC (perhaps due to reduced partial volume effect), markedly lower spatial distortion, and greater preservation of signal across the entire image volume.
While sensory activation is observed only in the single-shot image, this dataset also contains many false-positives; the activation patterns in the 4-shot datasets are highly specific to motor activity. These results, and resting-state fMRI results presented by Barry et al. [6], therefore support the advantages of multi-shot GRE methods for minimizing spatial distortion and maximizing resolution in 7T spinal cord fMRI. Multi-shot protocols are, however, more susceptible to time-varying fields, such as those caused by respiration, and rely on compensation for this effect both at the level of image reconstruction and time series analysis [5,6].
[1] Stroman PW, Wheeler-Kingshott C, Bacon M, et al. The current state-of-the-art of spinal cord imaging: methods. NeuroImage 2014;84:1070-1081.
[2] Cohen-Adad J. Functional Magnetic Resonance Imaging of the Spinal Cord: Current Status and Future Developments. Semin Ultrasound CT MRI 2017;38:176-186.
[3] Zhang B, Seifert AC, Kim JW, Borrello J, Xu J. 7 Tesla 22-Channel Wrap-Around Coil Array for Cervical Spinal Cord and Brainstem Imaging. Magnetic Resonance in Medicine 2017;78(4):1623-1634.
[4] Barry RL, Vannesjo SJ, By S, Gore JC, Smith SA. Spinal Cord MRI at 7T. NeuroImage 2018;168:437-451.
[5] Vannesjo SJ, Miller KL, Clare S, Tracey I. Spatiotemporal characterization of breathing-induced B0 field fluctuations in the cervical spinal cord at 7T. NeuroImage 2018;167:191-202.
[6] Barry RL, Smith SA, Dula AN, Gore JC. Resting state functional connectivity in the human spinal cord. Elife 2014;3:02812.
[7] Jenkinson M, Smith SM. A global optimisation method for robust affine registration of brain images. Medical Image Analysis, 5(2):143-156, 2001.
[8] Woolrich MW, Ripley BD, Brady M, Smith SM. Temporal Autocorrelation in Univariate Linear Modeling of FMRI Data. NeuroImage 2001;14(6):1370-1386.
[9] Kong Y, Jenkinson M, Andersson JLR, Tracey I, Brooks JCW. Assessment of physiological noise modelling methods for functional imaging of the spinal cord. NeuroImage 2012;60(2):1538-1549.