Wiki

Clone wiki

ACID - Artefact correction in diffusion MRI / Image cropping

Image cropping

Most SC dMRI sequences use reduced-FOV or ZOOM sequences which feature a small FOV in the phase-encoding direction (typically anterior/posterior) and a large FOV along the frequency-encoded direction (typically left/right) (see figure above). This means that the FOV covers a lot of structures outside the SC. As the first step of the pre-processing, it is recommended to crop the SC dMRI dataset, e.g. reduce the field of view (FOV), to exclude much of the surrounding tissue.

Cropping improves the performance of ECMOCO, as structures surrounding the spinal cord (bones, ligaments, etc.) can move independently from the spinal cord, which can result in incorrect estimation of transformation parameters for the spinal cord itself. On the other hand, cropping greatly reduces the size of the dMRI dataset and therefore accelerates all subsequent processing steps, especially the ECMOCO and model fitting. Note that large inter-image movements across the dMRI dataset might interfere with the cropping (i.e. the spinal cord moves out of the cropped FOV after large movement). In this case, ECMOCO should precede cropping. Also note that SC dMRI is typically acquired using reduced-FOV sequences, which already have a small FOV in the phase-encoding direction, so cropping is usually necessary only in the frequency-encoding direction.

The tool is available under: SPM -> Tools -> ACID -> Utilities -> Crop images

ECMOCO Background Fig. 3: Image cropping tool within the ACID toolbox

This tool allows the user to crop the dMRI images either automatically by specifying the midpoint (around which the image will be cropped) and the new matrix size, or manually by selecting the in-plane midpoint using crosshairs in the pop-up window (see image below). After specifying the cropped FOV based on the first image, the cropping will automatically be applied on all other selected images of the dMRI dataset.

The following input parameters have to be specified:

  • Data to be cropped: select your dMRI dataset. It must be in nifti format and can be 3d or 4d.
  • New image matrix: enter a 1x3 array specifying the desired matrix size of the cropped image in the x, y, and z directions, respectively. Enter a large number if you don't want to crop the image along a particular direction.
  • Midpoint in the axial plane: the indices of the midpoint voxel in the axial plane, around which the image will be cropped in the axial plane. If left unspecified (click OK without having entered a value), a window will pop up showing the specified middle slice of the first volume. In this window, you have to manually select the midpoint by moving the cursor to the desired point and pressing the left mouse button.
  • Prefix for output: prefix for the output cropped image

ECMOCO Background Fig. 2: Illustration of image cropping tool on a typical reduced-FOV axial DTI dataset with an image matrix of 356 x 80 x 10. After entering the matrix size of the cropped image and specifying the midpoint by entering a value or manually selecting it in the pop-up window (shown on the bottom left), the program will generate a cropped 4D dMRI dataset where the specified FOV is applied on all selected dMRI images.

Updated