Researchers and clinicians interested in simultaneous functional imaging of the spinal-cord and brainSimultaneous functional brain and spinal-cord MRI would be important for understanding how neural information is processed but is challenging due to poor $$$B_0$$$ homogeneity in the neck. This is exacerbated by the small size of the spinal-cord, which prompts a higher spatial resolution [1]. $$$B_0$$$ homogeneity can be improved by shimming [2], but the higher-order shims on most scanners are static and limited to second- or third-order, making it difficult to shim the brain and spinal-cord simultaneously. One approach has been to vary the shims dynamically, i.e. slice-specific linear shims and frequency offset between the two volumes [3], and perform reduced FOV imaging using a 2D slice-selective RF pulse to shorten the readout length [4]. There are problems with this approach, however. 1) Due to the (in-plane) symmetry of the neck, varying the linear shims has limited effect. 2) a 2D spinal-cord slice-selective RF pulse is too long and highly susceptible to off-resonance. 3) A volume-based shim trades-off homogeneity in one region for another, and may degrade image quality in the brain. Here, we propose simultaneous brain and spinal-cord functional imaging using: 1) slice-based shim + 2D imaging for the brain 2) volume-based shim + 2D slab-selective RF pulse + 3D imaging for the spinal-cord with EPI readout direction orthogonal to the RF pulse dimensions. The slice-based shim reduces trade-off in homogeneity [5]. The purpose of 3D imaging in the spinal-cord is to reduce the length of the RF pulse, and hence susceptibility to off-resonance. The orthogonal EPI readout direction allows aliasing to be prevented in the non-selective dimension of the RF pulse.

The static shims on our scanner are the second-order spherical harmonics ($$$xy$$$, $$$zy$$$, $$$zx$$$, $$$x^2-y^2$$$, and $$$z^2-(x^2+y^2)/2$$$), and the dynamic shims are the linear gradients ($$$x$$$, $$$y$$$, $$$z$$$) and a frequency offset $$$\Delta f$$$. We minimize the total off-resonance (in the sum-of-squares sense) over each voxel in a mask region over the brain and spinal-cord using these shims, where one set of static shims is used for the entire scan, and a separate set of dynamic shims for each slice and for the neck volume. The $$$zx$$$ and $$$zy$$$ coils were not used for the static shim and the shim was not used for the slices since they were (completely or mostly) redundant and resulted in ill-conditioning of the problem. Due to imperfections in the second-order shims, i.e. they do not generate fields that correspond exactly to the spherical harmonics, we modeled each one in terms of all the shims (static and dynamic), and calculated the shim amplitudes required to produce the desired field most accurately (in the sum-of-squares sense).

To test the excitation, we used a 2DFT readout with the following sequence parameters common to head and neck: matrix-size $$$64 \times 64$$$, readout bandwidth $$$BW_r = \pm 125$$$ kHz, echo-time $$$TE = 10$$$ ms, repetition-time $$$TR = 100$$$ ms. For the head, we used: slice thickness $$$\Delta z = 4$$$ mm, $$$FOV_{head} = 22$$$ cm, # slices $N_s = 6$. For the neck, we used: slab-thickness $$$D_z = 4$$$ cm, slab-width $$$D_y = 4$$$ cm, time-bandwidths $$$TB_z = TB_y = 4$$$, $$$FOV_x = 12$$$ cm, $$$FOV_y = 6$$$ cm, $$$FOV_z = 5$$$ cm, # z phase-encodes $$$N_z = 6$$$.

The dynamic shim improves the field in the brain but has limited effect in the neck (see Figure 1). Using the dynamic shim and a 2D EPI RF pulse, we achieved reasonable selectivity in the neck. We acquired a slightly larger FOV in z (factor of 1.25) to reduce aliasing effects. Ideally, the excited region would have constant width (and thickness), but this may not be the case due to off-resonance. However, since only the spinal-cord is of interest, imperfect excitation outside it in the frequency-encode direction is not problematic. To avoid aliasing in the frequency-encode direction, we increased the readout bandwidth and reconstructed a larger FOV, since the traditional method of applying an anti-aliasing filter does not work with ramp-sampling. Due to alternating between acquiring a 3D volume and other acquisitions, the time between each z phase-encode in the neck is not constant, i.e. the spins are more relaxed during the earlier phase encodes. This may be addressed by varying the flip-angles to yield the maximum total signal or interleaving head slices and neck phase-encodes to allow greater $$$T_1$$$ relaxation. We plan to apply this method to simultaneous brain + spinal cord fMRI using a sensory-motor task with GRAPPA-accelerated EPI.