Spinal cord smoothing as a preprocessing operation

SCT provides a spinal cord smoothing function that takes into account the curvature of the cord, preserving the boundary between the spinal cord and cerebrospinal fluid. This is useful in a variety of situations:

• It can be used to improve sensitivity of fMRI results while minimizing contamination from CSF

• It can also be used to obtain more reliable cord segmentations, because smoothing will sharpen the edge of the cord and will blur out possible artifacts at the cord/CSF interface.

Downloading the files for this tutorial

First, make sure that you have the following files in your working directory:

• single_subject/data/t1.nii.gz: An anatomical spinal cord scan in the T1 contrast.

• single_subject/data/t1_seg.nii.gz: 3D segmentation of the spinal cord, corresponding to the T1 image.

  You can get these files by downloading data_spinalcord-smoothing.zip.

Running the command

Next, open up your terminal and run the following command:

sct_smooth_spinalcord -i t1.nii.gz -s t1_seg.nii.gz

Input arguments:

• -i : The input image.

• -s : A spinal cord segmentation mask corresponding to the input image. This is needed as sct_smooth_spinalcord performs a 1D smoothing operation following the cord centerline (as opposed to the cortical surface smoothing in FreeSurfer, which is 2D).

Output files/folders:

• t1_smooth.nii.gz : The input image, smoothed along the spinal cord.

• warp_curve2straight.nii.gz : sct_smooth_spinalcord involves an intermediate straightening step, so this is the 4D warping field that defines the transform from the original curved anatomical image to the straightened image.

• warp_straight2curve.nii.gz : sct_smooth_spinalcord involves an intermediate straightening step, so this is the 4D warping field that defines the inverse transform from the straightened anatomical image back to the original curved image.

• straight_ref.nii.gz : The straightened input image produced by the intermediate straightening step. Can be re-used by other SCT functions that need a straight reference space.

• straightening.cache : SCT functions that require straightening will check for this file. If it is present in the working directory, straight_ref.nii.gz and the two warping fields will be re-used, saving processing time.

After smoothing, the apparent noise is reduced, while the cord edges are preserved. This may allow for the spinal cord to be segmented in cases where segmentation tools initially produce poor results.

However, you should always inspect the output and perform manual correction as necessary. This is especially critical when performing quantitative analyses (i.e. CSA calculation) across multiple subjects. If you perform smoothing on only some images in the dataset, you may introduce bias via over/under-segmenting that subset of images relative to the rest.

You can always compare the quality of the segmentation produced on the smoothed image by running the following commands:

# Second-pass segmentation using the smoothed anatomical image
sct_deepseg spinalcord -i t1_smooth.nii.gz -qc ~/qc_singleSubj