Introduction

One of the main challenges to obtaining high quality images of the spinal cord is the fact that the spinal cord is close to the lungs, which cause artifacts due to the difference in magnetic susceptibility between air and tissue. During respiration, the volume of air in the lungs varies, in turn causing magnetic field (B0) variations that are changing throughout respiration and lead to image artifacts¹. Previous studies introduced z-shimming for improving spinal cord acquisitions², however this approach is static in that each slice is assigned a unique Gz value, regardless of respiratory state. Other work introduced a method to account for respiratory-related B0 variations³, however, this work required the use of a dedicated external shim coil hardware and hence is not practical. We propose a novel solution, real-time z-shimming, to correct for both static and dynamic B0 variations during image acquisition.

Methods

We implemented real-time z-shimming within a 2D multi-gradient echo (MGRE) sequence (see the sequence diagram in Fig. 1) by adding “compensation” gradients along the z-axis, prior to every readout, with momentum equal to -Gz*TE, where Gz is the local field gradient along the slice-select direction (z) and TE is the echo time.

A single volunteer was scanned on a Siemens 3T scanner with a 64-channel head-neck coil and a 48-channel spine coil. A respiratory bellow (Siemens) was placed on the volunteer’s chest to inform on the respiratory state, via pressure readings (as shown in Fig. 1). A series of single-slice sagittal B0 maps were first acquired along the midline of the spinal cord, using a B0 mapping sequence (TR=7.2ms, TE1=2.46ms, TE2=4.93ms, matrix=96x64, image resolution=3x3x3mm³, 60 repetitions). From the B0 images, a time series of Gz maps was generated so that the respiration-induced Gz changes could be determined by examining the linear relationship between the time-varying Gz maps and the respiratory cycle^3,4. This results in a relationship between Gz(t) and pressure, P(t):

Gz(t) = Gz,static + c(z)*P(t) [Eq.1]

A short axial GRE scan (without shimming) was also acquired for the purpose of automatic spinal cord segmentation using SCT⁵, so that average Gz(t) values could be obtained for the spinal cord region. Those values were then quickly transferred to the scanner via ethernet socket and MGRE images with real-time z-shimming were acquired (TR=1000 ms, 6 echoes with 3ms inter-echo spacing, TE1=2.5ms, BW = 600 Hz/px, 20 slices, matrix=128x68, image resolution=2.2x2.2x3mm³). During the shimmed acquisition, the pressure from the respiration bellow was sampled prior to each slice acquisition and the appropriate compensation gradient was calculated (according to Eq. 1).

Results

Real-time z-shimming was found to recover signal loss in the spinal cord. In Fig. 2 we show sagittal and axial views of the spinal cord with and without real-time z-shimming. The red arrow indicates areas in which real-time z-shimming recovered some of the signal that is lost due to respiration-induced B0 field inhomogeneities.

Discussion and Conclusions

We introduced a novel approach that combines z-shimming with real-time update from respiration trace, for compensating both static and dynamic B0 variations in the spinal cord. Our approach does not require additional software, making its translation to other centers straightforward. The proposed method will help make MRI techniques that are sensitive to magnetic field inhomogeneities, such as DTI, functional MRI and spectroscopy more feasible in the spinal cord. We expect the gain to be even larger at ultra-high field, which will be the purpose of follow-up studies.

Acknowledgements

Funded by the Canada Research Chair in Quantitative Magnetic Resonance Imaging [950-230815], the Canadian Institute of Health Research [CIHR FDN-143263], the Canada Foundation for Innovation [32454, 34824], the Fonds de Recherche du Québec - Santé [28826], the Fonds de Recherche du Québec - Nature et Technologies [2015-PR-182754], the Natural Sciences and Engineering Research Council of Canada [RGPIN-2019-07244], the Canada First Research Excellence Fund (IVADO and TransMedTech), the Courtois NeuroMod project and the Quebec BioImaging Network [5886, 35450].