Reorient and resample in generic SCT processing

Marcella_Lagana · June 5, 2020, 9:01pm

Why reorientation and resampling are performed on sagittal T1w and T2w images before any other processing?

I mean the following lines:

Reorient to RPI and resample to 1mm iso (supposed to be the effective resolution)

sct_image -i ${file_t1}.nii.gz -setorient RPI -o ${file_t1}_RPI.nii.gz
sct_resample -i ${file_t1}_RPI.nii.gz -mm 1x1x1 -o ${file_t1}_RPI_r.nii.gz

Reorient to RPI and resample to 0.8mm iso (supposed to be the effective resolution)

sct_image -i ${file_t2}.nii.gz -setorient RPI -o ${file_t2}_RPI.nii.gz
sct_resample -i ${file_t2}_RPI.nii.gz -mm 0.8x0.8x0.8 -o ${file_t2}_RPI_r.nii.gz

For examples in

github.com

sct-pipeline/spine-generic/blob/master/processing/process_data.sh

#!/bin/bash
#
# Process data.
#
# Usage:
#   ./process_data.sh <SUBJECT>
#
# Authors: Julien Cohen-Adad

# The following global variables are retrieved from parameters.sh but could be
# overwritten here:
# PATH_QC="~/qc"

# Uncomment for full verbose
set -v

# Immediately exit if error
set -e -o pipefail

# Exit if user presses CTRL+C (Linux) or CMD+C (OSX)

This file has been truncated. show original

or https://github.com/sct-pipeline/hagiwara-ms/blob/master/processing/process_data.sh

Thanks!
Marcella

jcohenadad · June 7, 2020, 2:06pm

Hi @Marcella_Lagana,

Reorienting the image has no impact on the data matrix (only the header is affected). This is just done to ensure consistent orientation for subsequent processes (in case some people used another orientation than what was prescribed).

Resampling the images is done to ensure consistent image resolution for subsequent processes, e.g. cord segmentation, registration to template, etc. In some vendors, e.g. GE scanners, the output image matrix (resulting from the inverse FFT) is twice as large as the acquisition matrix. So, for an effective acquisition resolution of 1x1x1mm, the reconstructed image is 0.5x0.5x1, which is confusing, takes more space on the disk, and makes the analysis longer. So for consistency we just resample everything to the nominal (expected) resolution.

Cheers,
Julien