DTI

All things diffusion MRI

Intro

Diffusion imaging is a magnetic resonance imaging (MRI) technique that measures the movement of water molecules in tissues, providing insights into tissue microstructure and organization.

DWI vs DTI

• Diffusion-weighted MRI (DWI):
  This type of MRI uses magnetic field gradients to create an image that is sensitive to the diffusion of water molecules.
• Diffusion Tensor Imaging (DTI):
  DTI builds upon DWI by measuring the direction and magnitude of water diffusion, allowing for the visualization of neural pathways and fiber tracts. Makes some assumptions / approximations that may not hold true.

Simple DTI analysis

Here is a script that uses FSL to run a DTI analysis.

Files need to be converted from source files (DICOM or PAR/REC for example) into NIFTI and BVEC/BVAL. This can be done using dcm2niix

The pipeline requires two inputs: main DTI file basename in one phase direction; a DTI nifti file acquired in the opposite phase direction

usage:

fsldti_pipeline <main_data_name> <b0oppositephase_name>.nii.gz

[note: main_data_name doesn't have a suffix. Just the basename]

This pipeline also assumes you have matlab, fsl, and a version of eddy_cuda working.

Diffusion Imaging with Python (DIPY)

Videos:
https://www.youtube.com/c/DiffusionImaginginPython/videos

QSIPrep

Like fmriprep for DTI: QSIPrep

NODDI

What is it? Here is a good paper explaining the concept and clinical applications: https://www.sciencedirect.com/science/article/pii/S0165027020303319

Basically:
NODDI stands for Neurite Orientation Dispersion and Density Imaging. It’s a diffusion MRI technique used to estimate the microstructural complexity of axons and dendrites in the brain. This method provides more specific markers of brain tissue microstructure compared to traditional indices from diffusion tensor imaging (DTI). Usually requires more than one shell.

AMICO

https://github.com/daducci/AMICO/wiki/Installation

pip install dmri-amico

Example:

In Python:

import amico
amico.setup()
amico.util.fsl2scheme("derivatives/qsirecon/sub-Pilot01/dwi/sub-Pilot01_space-T1w_desc-preproc_fslstd_dwi.bval", "derivatives/qsirecon/sub-Pilot01/dwi/sub-Pilot01_space-T1w_desc-preproc_fslstd_dwi.bvec")

ae = amico.Evaluation("derivatives/qsirecon", "sub-Pilot01/dwi")
ae.load_data( "sub-Pilot01_space-T1w_desc-preproc_fslstd_dwi.nii.gz", "sub-Pilot01_space-T1w_desc-preproc_fslstd_dwi.scheme", mask_filename="sub-Pilot01_space-T1w_desc-preproc_fslstd_mask.nii.gz", b0_thr=0 )

ae.set_model("NODDI")
ae.generate_kernels()

ae.load_kernels()
ae.fit()
ae.save_results()

As a Python Script you can run:

import amico
import os
import sys, getopt

def main(argv):
   dwifile = ''
   maskfile = ''
   opts, args = getopt.getopt(argv,"hd:m:",["dfile=","mfile="])
   for opt, arg in opts:
      if opt == '-h':
         print ('test.py -d <dwifile> -m <maskfile>')
         sys.exit()
      elif opt in ("-d", "--dwifile"):
         dwifile = arg
      elif opt in ("-m", "--maskfile"):
         maskfile = arg

   basename=os.path.splitext(dwifile)[0]
   schemefile=basename+'.scheme'
   bvalfile=basename+'.bval'
   bvecfile=basename+'.bvec'
   print ('Bval file is ', bvalfile)
   print ('Bvec file is ', bvecfile)
   print ('Scheme file is ', schemefile)
   print ('Dwi file is ', dwifile)
   print ('Mask file is ', maskfile)

   amico.setup()
   amico.util.fsl2scheme(f"{bvalfile}", f"{bvecfile}")

   ae = amico.Evaluation(".", ".")
   ae.load_data( f"{dwifile}", f"{schemefile}", mask_filename=f"{maskfile}", b0_thr=0 )

   ae.set_model("NODDI")
   ae.generate_kernels()

   ae.load_kernels()
   ae.fit()
   ae.save_results()

if __name__ == "__main__":
   main(sys.argv[1:])

Matlab Toolbox

http://mig.cs.ucl.ac.uk/index.php?n=Tutorial.NODDImatlab

The toolbox requires:

• Matlab with the optimization toolbox
• NIFTI Matlab library: nifti_matlab

First run the simple DTI script from before (https://github.com/WeberLab/Misc/blob/main/dtipipeline.sh) script to get a 'topped-up' DTI and brain mask to use.

Then convert to Analyze (AFNI):

3dcalc -a eddy_corrected_data.nii -expr 'a' -prefix eddy_corrected_data.hdr
3dcalc -a DTI_topupapplied_brain_mask.nii -expr 'a' -prefix DTI_topupapplied_brain_mask.hdr

In Matlab:

addpath(genpath('/home/weberam2/matlab/NODDI_tool')) % or whever the NODDI from Github is
addpath(genpath('/home/weberam2/matlab/nifti_matlab')) % or whever your nifti_matlab is
CreateROI('eddy_corrected_data.hdr', 'DTI_topupapplied_brain_mask.hdr', 'NODDI_roi.mat');
protocol = FSL2Protocol('DTI.bval', 'DTI.bvec');
noddi = MakeModel('WatsonSHStickTortIsoV_B0');
batch_fitting('NODDI_roi.mat', protocol, noddi, 'FittedParams.mat', 8);
SaveParamsAsNIfTI('FittedParams.mat', 'NODDI_roi.mat', 'DTI_topupapplied_brain_mask.hdr', 'outputname')

Python Version

https://github.com/dicemt/DTI-NODDI

dHCP DWI Pipeline

https://git.fmrib.ox.ac.uk/matteob/dHCP_neo_dMRI_pipeline_release

Tractrography

ractography is a 3D modeling technique in neuroscience that visually represents white matter pathways within the brain using data collected from diffusion MRI. This technique provides a non-invasive way to explore the brain’s structural connectivity.

Probtrackx (FSL)

https://fsl.fmrib.ox.ac.uk/fsl/fslwiki/FDT/UserGuide

https://fsl.fmrib.ox.ac.uk/fslcourse/2019_Beijing/lectures/FDT/fdt2.html

MRTRIX3

see add link

FSL's DTIFIT

Radial Diffusivity

RD = (L2 + L3) / 2

$FSLDIR/bin/fslmaths ${dtifit_folder}/${subj}*_L2.nii.gz –add ${dtifit_folder}/${subj}*_L3.nii.gz \\
       -div 2 ${parentDirectory}/RD/${subj}_RD.nii.gz