DTI pre-processing for the DRC

From CMIC
Jump to: navigation, search

Diffusion Tensor Imaging depicts the anisotropy of tissue.

The acquisition protocol includes 6 or more diffusion weighted images (DWIs), one or more B null image (B0), all encoded into a 4D nifti file, its corresponding bval-bvec pair, a structural T1 image, an optional brain mask, and and optional field maps (FM) magnitude and phase that are used for susceptibility correction.

Within the DRC network, the nipype script to process the diffusion images is perform_dti_processing_drc_protocol.py. In the DRC, the DWIs acquisition is repeated to increase SNR, the acquisition is identified in the databases with a 4 or 5 digit number, the midas code MIDAS_code. As a consequence, this is the only required input argument for the script:

perform_dti_processing_drc_protocol.py -m <MIDAS_code> -o output_dir

The rest of the optional options are described below:
<pre>
usage: perform_dti_processing.py [-h] -m midas_code [--rot rot] [--etd etd] [--ped [ped]]
                                 [--rigid]

Perform Diffusion Model Fitting with pre-processing steps. Mandatory input 
is the midas_code. Values to use for the susceptibility
parameters: ## DRC ## (--ped=-y --etd=2.46 --rot=34.56) and ## 1946 ##
(--ped=-y --etd=2.46 --rot=25.92). Note that these values are indicative.

optional arguments:
  -h, --help            show this help message and exit
  -m midas_code 
                        midas code to use as input identifier, 4-5 digit
  -o output_dir, --output_dir output_dir
                        Output directory containing the registration result
                        Default is a directory called results
  -g, --graph           Print a graph describing the node connections
  --rot rot             Diffusion Read-Out time used for susceptibility
                        correction Default is 34.56
  --etd etd             Echo Time difference used for susceptibility
                        correction Default is 2.46
  --ped [ped]           Phase encoding direction used for susceptibility
                        correction (x, y or z) --ped=val form must be used for
                        -ve indicesDefault is the -y direction (-y)
  --rigid               Only use rigid registration for DWI (no eddy current
                        correction)

The outputs are gathered in the -o argument (e.g. output_dir. All output files share a similar prefix in their filename, corresponding to the basename of the input DWI file. It usually corresponds to a subject identifier (e.g. subject_id

Here is a list of the principal outputs and their corresponding description:

_corrected_dwi: pre-processed merged diffusion weighted images in a 4D nifti file. (motion, susceptibility, etc)
_corrected_dwi.bval/bvec: corresponding bvalues and gradient vector direction files
_average_b0: the average B-Null image
_dwi_to_b0_rotation.png / _interslice_ncc.png : QC plots describing respectively the subject motion and the signal dropouts throughout the DWI scans
_tensors.nii.gz : output fitted tensor image file, in a 4D format with 6 degrees as a 4th dimension, from which the diffusion based biomarker maps are estimated
_famap : the fractional anisotropy map (3D)
_mdmap : the mean diffusivity map (3D)
_v1map : the first eigen vector component map (3D x 3)
_rgbmap : the directional colour coded RGB FA map.
_mask : mask used to crop background during processing, in the diffusion space
_T1/T2_to_B0.txt : affine transformations between T1/T2 and the average B0 image (with the T1 as reference)