The spinal cord is a particularly challenging structure to image with MRI, due to its small cross-sectional anatomy and its location deep inside the body. For accurate depiction of small structures within the cord, the SNR benefit of higher background field strengths is desirable. However, several technical difficulties limit the implementation of ultra-high field spinal cord imaging. Especially problematic are static and dynamic B₀ field variations, which cause image artifacts in the form of signal loss, ghosting and distortion.

A special case of dynamic B₀ field variations are fields induced by the change in lung volume during the breathing cycle. Breathing-related field fluctuations on the order of 10-20Hz have been reported to significantly deteriorate quality of T2*-weighted images in brain imaging at 7T¹. Due to the spine being closer to the thorax and the lungs, the breathing-related fields will expectedly have an even stronger impact on spinal cord imaging. There is, however, to date no study showing the extent of breathing-related B₀ field variations in the spinal cord at ultra-high field.

Here we investigate the magnitude and spatial profile of breathing-induced field fluctuations in the cervical spinal cord at 7T. We further explore the feasibility of actively compensating for the induced field changes by adjusting the shim settings according to breathing state.

Spinal cord imaging of two healthy volunteers (male) was performed on a Siemens Magnetom 7T system, using a volume transmit, 16-ch receive cervical spine coil (Quality Electrodynamics). Fast low-resolution field maps (FOV=152x152mm², 2mm isotropic resolution, 11 sagittal slices, TR=80ms, TE1= 4.08ms, ΔTE=1.02ms, acq. time 12s) of the cervical spinal cord were acquired during breath-holds, either in an expired or inspired breathing state. The field map acquisitions were repeated three times per condition in each subject. The field difference between each pair of expired/inspired field maps was quantified within a manually drawn mask covering the spinal cord.

In one subject, a shim correction was implemented to compensate for the breathing-induced fields. Initially, expired/inspired field maps were acquired with a fixed shim field setting. Up to 2^nd-order spherical harmonic shim fields were then fit to the field difference within the spinal cord mask, assuming expiration as reference state. The calculated correction shim settings were subsequently applied during the inspired condition of a new field map acquisition pair within the same scan session. The compensated acquisition pair was repeated twice.

A reproducible pattern of breathing-related field changes was observed in both subjects (Fig. 1). The strength of the breathing-induced fields was highly variable along the length of the spinal cord, reaching up to 140Hz and 100Hz at the level of C7 for the two different subjects respectively (Fig. 2). At the level of C1 the difference was down to around 10Hz for both subjects.

The field difference (ΔB₀) was greatly reduced by using 2^nd-order shim compensation (Figs. 2 and 3). With compensation, ΔB₀ maximally peaked at around 30Hz, and the variance along the cord was less. The mean and the standard deviation of ΔB₀ over all voxels within the spinal cord mask were reduced by almost a factor of three as compared to uncompensated acquisitions. Remaining offsets are likely to be attributed to variations in the depth of the breath-hold between acquisitions.

We here measured breathing-related field changes of up to 140Hz in the cervical spinal cord at 7T. This agrees well with a previous study by Verma et al reporting breathing-related B₀ field variations of around 70Hz at the level of C7 at 3T². Field variations on this order of magnitude will be detrimental for any imaging sequences with long TE, including functional MRI, T2*-weighted imaging and susceptibility-weighted imaging. In single-shot EPIs the field variations will manifest as large image shifts, whereas structural imaging will suffer heavily from ghosting. Through-slice field gradients will furthermore contribute to dephasing and signal loss. There is thus a strong need for correction strategies addressing the B₀ field variations.

We demonstrated a proof-of-principle shim correction indicating that it would be feasible to greatly reduce the level of fluctuations by adjusting the shim settings according to the breathing state. Real-time shim updating to compensate for breathing-induced field variations has previously been shown to improve brain imaging at 7T^3,4. The results here suggest that the benefit of such compensation would be even greater for spinal cord imaging, especially for applications relying on T2*-weighted sequences, such as functional imaging.