Ten Key Insights into the Use of Spinal Cord fMRI

Abstract

A comprehensive review of the literature-to-date on functional magnetic resonance imaging (fMRI) of the spinal cord is presented. Spinal fMRI has been shown, over more than two decades of work, to be a reliable tool for detecting neural activity. We discuss 10 key points regarding the history, development, methods, and applications of spinal fMRI. Animal models have served a key purpose for the development of spinal fMRI protocols and for experimental spinal cord injury studies. Applications of spinal fMRI span from animal models across healthy and patient populations in humans using both task-based and resting-state paradigms. The literature also demonstrates clear trends in study design and acquisition methods, as the majority of studies follow a task-based, block design paradigm, and utilize variations of single-shot fast spin-echo imaging methods. We, therefore, discuss the similarities and differences of these to resting-state fMRI and gradient-echo EPI protocols. Although it is newly emerging, complex connectivity and network analysis is not only possible, but has also been shown to be reliable and reproducible in the spinal cord for both task-based and resting-state studies. Despite the technical challenges associated with spinal fMRI, this review identifies reliable solutions that have been developed to overcome these challenges.

Keywords: analysis; animal; fMRI; human; methods; review; spinal cord.

Figure 1

Histogram of the number of papers published per year on spinal fMRI, since its inception in 1996.

Figure 2

Flow chart of the literature review process.

Figure 3

Chart of the distribution of spinal fMRI papers published to date, broken down by type, methods, study groups, and primary topic. HASTE—half-Fourier single-shot turbo spin echo; GE-EPI—gradient-echo with echo-planar spatial encoding; MS—multiple sclerosis; SCI—spinal cord injury; FMS—fibromyalgia syndrome.

Figure 4

Comparison of image quality obtained with gradient-echo EPI and spin-echo HASTE sequences for spinal fMRI. Gradient-echo EPI images were acquired in contiguous axial slices (A) and were reformatted into sagittal views (B) for comparison with spin-echo (HASTE) images acquired in sagittal planes (C). Selected axial slices are shown for comparison, and the slice positions are indicated relative to the caudal medulla (top slice). Images were acquired at 3 tesla with a Siemens MAGNETOM Trio at Queen’s University, and used as examples for this review.

Figure 5

Simulations of data analysis concepts for the purpose of this review article. The left panel (a) shows a model paradigm, P (black), fit to a simulated signal, S (red), consisting of a 1% BOLD signal change from a baseline value of 100, plus random noise. The results of a GLM fit are shown: S = 0.97 P + 100.0. The correlation between the paradigm and signal is shown as R = 0.978. The middle panel (b) shows the effect of sorting both the paradigm values and the signal values to order the signal values in increasing order. The results of the GLM fit and correlation are unchanged. The right-most panel (c) shows a plot of signal values versus paradigm values, and a linear fit between them. The fit and correlation results are again the same.