Dynamic per slice shimming for simultaneous brain and spinal cord fMRI

Abstract

Purpose: Simultaneous brain and spinal cord functional MRI is emerging as a new tool to study the central nervous system but is challenging. Poor B₀ homogeneity and small size of the spinal cord are principal obstacles to this nascent technology. Here we extend a dynamic shimming approach, first posed by Finsterbusch, by shimming per slice for both the brain and spinal cord.

Methods: We shim dynamically by a simple and fast optimization of linear field gradients and frequency offset separately for each slice in order to minimize off-resonance for both the brain and spinal cord. Simultaneous acquisition of brain and spinal cord fMRI is achieved with high spatial resolution in the spinal cord by means of an echo-planar RF pulse for reduced FOV. Brain slice acquisition is full FOV.

Results: T2*-weighted images of brain and spinal cord are acquired with high clarity and minimal observable image artifacts. Fist-clenching fMRI experiments reveal task-consistent activation in motor cortices, cerebellum, and C6-T1 spinal segments.

Conclusions: High quality functional results are obtained for a sensory-motor task. Consistent activation in both the brain and spinal cord is observed at individual levels, not only at group level. Because reduced FOV excitation is applicable to any spinal cord section, future continuation of these methods holds great potential.

Keywords: dynamic shimming; fMRI; reduced FOV; simultaneous brain and spinal cord fMRI; spinal cord.

FIGURE 1

A, Masked adjusted sagittal field map after the first‐step shim calculations (1). Dashed line at z = 9 cm indicates the sagittal slice location z₀. Red dashed lines indicate ±4 cm bounds about z₀. All lines (except red) indicate the slice locations. B, Blue dots represent off-resonance values within the masked spinal cord and within bounds. Red line fits those off-resonance values quadratically. Black line is tangent to the quadratic at z₀. This quadratic fit is repeated for each and every slice location z₀ over brain and spinal cord

FIGURE 2

A, Sagittal gradient-echo image with red lines indicating the slice position; spinal slices are centered around the C5/C6 vertebrae. B, Masked sagittal field map using auto-shim. Masks remove tissue of no interest. Masked sagittal field map serves as input to dynamic shim calculation

FIGURE 3

A, Axial gradient-echo images of brain and spinal cord corresponding to slice locations in Figure 2A. B, Masked axial field maps o using auto-shim. C, Shimmed axial field maps o + Es via (1) (5) (7). D, Field maps measured after per slice shimming, for verification only, are in close agreement with calculated maps in panel C. Bottom row (a*) (b*) (c*) (d*) shows zoomed images of slices indicated by red squares from panels A–D

FIGURE 4

In-plane mean and standard deviation of off-resonance vs slice for: B, axial field maps o using auto-shim (blue), C, calculated shimmed axial field maps o + Es (red), D, measured axial field maps after proposed per slice shimming (green). Solid line segments (within B, C, D) correspond to mean off-resonance calculated from field maps in Figure 3B–D. Shaded bands B–D correspond to in-plane standard deviation of off-resonance. Average standard deviation over all slices: B, 29.6 Hz, C, 19.2 Hz, D, 12.9 Hz

FIGURE 5

Pulse sequence diagrams for brain and spinal cord slices. Spinal cord slices used an echo-planar RF pulse: length T = 18.6 ms, time-bandwidth product in the frequency-encode direction TB_freq = 2, time-bandwidth product in the phase-encode direction TB_phase = 4, excitation width D_phase = 6 cm, and alias distance R_phase = 40 cm. For each slice there is a distinct set of linear x, y, & z shims and center-frequency offset Δf-shim determined dynamically. (1) The two sequences are combined within a single TR, first scanning entire brain and then entire spinal cord. TE = 30 ms throughout

FIGURE 6

Comparison of the four shimming methods illustrated by: A, T2*-weighted images, B, corresponding measured field maps. Methods are: column 1: standard static linear and zero-order auto-shim, 2: static second through zero-order shim over both brain and spinal cord, 3: static second through zero-order shim on spinal cord only, 4: proposed dynamic per slice shim is the only method illustrated that incorporates field- map masking. Columns 1 & 2 exhibit image distortion with signal loss in the spinal cord. Column 3 shows nominal performance only in the spinal cord. Column 4 performs better overall. Spinal cord slice 24 in columns 1 & 2 and both brain slices in column 3 contain wrapped field estimates. Wrapping indicates out-of-range field map values, which means greater B₀ inhomogeneity. Dynamic per slice shim produces the most homogeneous B₀ field maps

FIGURE 7

A, GRASS images of brain, B, corresponding T2*-weighted images acquired by dynamic per slice shimming. Simultaneously acquired T2* spinal cord images are shown in Figure 8B

FIGURE 8

A, GRASS images of spinal cord, B, corresponding T2*‐weighted images acquired by dynamic per slice shimming. Simultaneously acquired T2* brain images are shown in Figure 7B. There is great resemblance between these GRASS and T2* images, suggesting that dynamic shim is effective in both brain and spinal cord regions

FIGURE 9

A, Brain and B, spinal cord fMRI activation maps from a fist-clenching task. For each subject and group activation map, axial and sagittal slices are shown. Level is indicated by blue lines; spinal cord segments C6 through T1 are illustrated. In brain slices, robust bilateral activation is observed in the motor areas and cerebellum (p < 1e−6). Spinal cord map threshold is p < 0.02 for subjects S1-S9. Group map threshold is Z > 2.3 with corrected cluster threshold p < 0.01. C, Activation time series, for subject S9, by averaging voxels in slice 25 (spinal cord) having Z > 4.5. High correlation with task is observed between measured and ideal time series

FIGURE 10

Linear z-shim and offset Δf-shim calculated from sagittal field maps during (blue) a 3 s fully inhaled state and (red) a 3 s fully exhaled state. Four maps are collected for each state. A, Optimal z-shim does not vary greatly, between states, both in the brain and spinal cord. Maximum z-shim difference between states is insubstantial: 0.016 mT/m corresponding to 2.73 Hz difference across the 4 mm slice thickness. B, Δf-shim accounts for frequency offset generated by the z-shim. While z-shim is insensitive to breathing, maximum difference of Δf-shim between the two breathing states is 70 Hz