Fig 1.
(A) The axial slice stacks of the 3D spoiled multi-echo gradient-echo (ME-GRE) sequence (indicated in green) and the diffusion MRI sequence (indicated in yellow), overlaid on the sagittal T2-weighted image. The field of view of the ME-GRE scan was set such that its 6th most rostral slice (slice #15) aligned with the maximum width of the spinal cord as observed in the sagittal T2-weighted scan (slice highlighted in red across all images). The slice stack of the ME-GRE scan covers the lumbosacral enlargement (LSE) and the entire conus medullaris (CM), while that of the diffusion MRI is smaller and centered at the 9th most rostral slice (slice #12) of the ME-GRE scan. (B) Corresponding axial slices of the ME-GRE scan. (C) Corresponding axial slices of the maps of diffusion tensor imaging metrics, including fractional anisotropy (FA), mean diffusivity (MD), axial diffusivity (AD), and radial diffusivity (RD). Axial slices are displayed in rostral (top left) to caudal (bottom right) direction.
Fig 2.
Adjustment for the individual conus medullaris (CM) length (60, 50, 40, and 30 mm for the displayed cases).
The acquired slices are illustrated by rectangular gray boxes (slice thickness of 5 mm). First, two image-based neuroanatomical landmarks (LSE, defined here as the slice with the largest cross-sectional GM area, and the CMtip) are determined in each subject (indicated by asterisks). Then, the space between these landmarks are divided into 5 segments of equal thickness (resulting in segment thicknesses of 12, 10, 8, and 6 mm). The segments, displayed as colored spinal cord regions, are defined such that one segment (light blue) is centered at the LSE landmark (segment LSE), another one at the tip of the spinal cord (segment LSE-5), and the space between them is covered by four segments. The spinal cord rostral to the LSE landmark is also divided into segments using the same segment thickness. Average values are extracted within each segment as a weighted average of the slice-wise values, where the weights represent the spatial contribution of each slice to the segment. If a value for a slice which contributes more than 25% to the segment was not available, the value for that segment was not calculated.
Fig 3.
Curves of slice-wise cross-sectional area (CSA) of the spinal cord (SC) and gray matter (GM), obtained by three different raters in two subjects, with and without applying moving window averaging across 3 adjacent slices.
The LSE landmarks, as determined by the raters, are indicated by asterisks. The benefit of using the slice with the largest GM CSA (GMmax,mw) as LSE landmark, as opposed to the slice with the largest SC CSA (SCmax,mw), is evident in Subject 1, with no inter-rater variability for the GMmax,mw slice. Subject 2 demonstrates the advantage of using moving window averaging: There is no inter-rater variability in GMmax,mw and SCmax,mw slice after applying moving window averaging.
Fig 4.
Inter-subject mean (solid line) and standard deviation (shaded area) of the cross-sectional area (CSA) of the spinal cord (SC) and gray matter (GM), computed across 10 healthy volunteers, when aligning the individual slice stacks at the slice either with the largest SC CSA (SCmax,mw, in red) or GM SCA (GMmax,mw, in blue), without adjusting for the length of the conus medullaris.
The inter-subject variability at caudal locations was slightly lower when aligning at the GMmax,mw slice, as seen by the smaller width of the blue shaded areas.
Table 1.
Inter-subject variability of cross-sectional area measurements.
Fig 5.
Illustration of intra- and inter-rater variability of manual spinal cord (SC) and gray matter (GM) segmentations within a representative subject, i.e., a subject whose intra- and inter-rater variability values are close to the mean values reported in Table 2.
SC and GM segmentations are shown on alternating slices for display purposes. For the displayed SC and GM segmentations, the coefficient of variation (CV) values of the corresponding SC cross-sectional areas (CSA) were (intra- vs. inter-rater) 0.8% vs. 3.7% (slice 17), 2.4% vs. 3.2%, 1.2% vs. 9.9%, 4.8% vs. 12.3%, and 6.8% vs. 6.7% (slice 9). The CV values of the corresponding GM CSA were (intra- vs. inter-rater) 1.8% vs. 9.1% (slice 16), 1.5% vs. 9.1%, 2.4% vs. 8.9%, 6.5% vs. 10.9%, and 3.6% vs. 9.5% (slice 8).
Table 2.
Slice-wise intra- and inter-rater reliability of cross-sectional area measurements (n = 10; 5 healthy volunteers and 5 patients with spinal cord injury).
Table 3.
Scan-rescan reliability of cross-sectional area measurements (n = 10 healthy volunteers).
Table 4.
Scan-rescan reliability of fractional anisotropy and mean diffusivity values (n = 10 healthy volunteers).
Table 5.
Scan-rescan reliability of axial and radial diffusivity values (n = 10 healthy volunteers).