Structural connectivity dMRI

Example workflow for constructing strutural connectivity (Human connectome project: Single subject)

This tutorial was created by Joan Amos.

Email: Github: @Joanone


The steps used for this tutorial were referenced from:

Data Description


The single subject data used in this tutorial has been preprocessed and was downloaded from:

Download demo data:!AjZJgBZ_P9UO8nWvAFwQyKQnrroe?e=6qmRlQ - Diffusion data!AjZJgBZ_P9UO8nblYQyUVsibqggs?e=mkwLpQ - Structural data

Required structural preprocessed input files

aparc+aseg.nii.gz T1w_acpc_dc_restore_brain.nii.gz

Required diffusion preprocessed input files

bvals bvecs data.nii.gz

Install Neurodesk on windows and mount external storage on your host computer


N/B: Constructing the structural connectivity using dMRI HCP data is computationally intensive. Thus, ensure you have sufficient disk space (>100GB) and RAM size (16, 32GB)

Open the powershell terminal and run:

docker run --shm-size=1gb -it --privileged --name neurodesktop -v C:/neurodesktop-storage:/neurodesktop-storage -v D:/moredata:/data -p 8080:8080 -h neurodesktop-20220222 vnmd/neurodesktop:20220222

N/B: The folder created in this tutorial was tagged “Test”

Open a terminal in neurodesk and run:


Activate mrtrix3 software in the neurodesk terminal

ml mrtrix3/3.0.3

N/B: The advantage neurodesk offers is the version of software can be selected from a range of others, which caters for reproducibility. The mrtrix3 (3.0.3) version was used in this tutorial.

Step 1: Further pre-processing

Extract data.nii.gz to enable memory-mapping. The extracted files are about 4.5GB:

gunzip -c data.nii.gz > data.nii; 
mrconvert data.nii DWI.mif -fslgrad bvecs bvals -datatype float32 -stride 0,0,0,1 -force -info;
rm -f data.nii

Perform mrconvert:

dwibiascorrect ants DWI.mif DWI_bias_ants.mif -bias bias_ants_field.mif -force -info;

Extract the response function. Uses -stride 0,0,0,1:

dwi2response dhollander DWI_bias_ants.mif response_wm.txt response_gm.txt response_csf.txt -voxels RF_voxels.mif -force; 

dwiextract DWI_bias_ants.mif - -bzero | mrmath - mean meanb0.mif -axis 3 -force -info

Generate mask:

dwi2mask DWI_bias_ants.mif DWI_mask.mif -force -info;

Generate Fibre Orientation Distributions (FODs):

dwi2fod msmt_csd DWI_bias_ants.mif response_wm.txt wmfod.mif response_gm.txt gm.mif  response_csf.txt csf.mif -mask DWI_mask.mif -force -info;

Perform normalization:

mtnormalise wmfod.mif wmfod_norm.mif gm.mif gm_norm.mif csf.mif csf_norm.mif -mask DWI_mask.mif -check_norm mtnormalise_norm.mif -check_mask mtnormalise_mask.mif -force -info

Generate a 5 tissue image:

5ttgen fsl T1w_acpc_dc_restore_brain.nii.gz 5TT.mif -premasked 

Convert the B0 image:

mrconvert meanb0.mif mean_b0.nii.gz

Activate the fsl and afni softwares in the neurodesk terminal:

ml fsl/6.0.3
ml afni/21.0.0

Use “fslroi” to extract the first volume of the segmented dataset which corresponds to the Grey Matter Segmentation:

fslroi 5TT.nii.gz 5TT_vol0.nii.gz 0 

Use “flirt” command to coregister the two datasets:

flirt -in  mean_b0.nii.gz -ref  5TT_vol0.nii.gz -interp nearestneighbour -dof 6 -omat diff2struct_fsl.mat

Convert the transformation matrix to a format readble by MRtrix:

transformconvert diff2struct_fsl.mat mean_b0.nii.gz 5TT.nii.gz flirt_import diff2struct_mrtrix.txt

Coregister the anatomical image to the diffusion image:

mrtransform 5TT.mif -linear diff2struct_mrtrix.txt -inverse 5TT_coreg 

Create the seed boundary which sepearates the grey from the white matter. The command “5tt2gmwmi” denotes (5 tissue type(segmentation) to grey matter/white matter interface):

5tt2gmwmi 5TT_coreg.mif gmwmSeed_coreg.mif

Step 2: Tractogram construction

The probabilistic tractography which is the default in MRtrix is used in this tutorial. The default method is the iFOD2 algorithm. The number of streamlines used is 10 million, this was chosen to save computational time:

tckgen -act 5TT_coreg.mif -backtrack -seed_gmwmi gmwmSeed_coreg.mif -nthreads 8 -minlength 5.0 -maxlength 300 -cutoff 0.06 -select 10000000 wmfod_norm.mif tracks_10M.tck -force

Step 3: SIFT2 construction

The generated streamlines can be refined with tcksift2 to counterbalance the overfitting. This creates a text file containing weights for each voxel in the brain:

tcksift2 -act 5TT_coreg.mif -out_mu sift_mu.txt -out_coeffs sift_coeffs.txt -nthreads 8 tracks.tck wmfod_norm.mif sift_1M.txt -force

Step 4: Connectome construction

In constructing the connectome, the desikan-killany atlas which includes the cortical and sub-cortical regions (84 regions) was used.

Copy the FreeSurferColorLUT.txt file from the ml freesurfer 7.2.0 singularity container to the subject’s folder

cp /opt/freesurfer-7.2.0/FreeSurferColorLUT.txt /data/Test/100307

Copy the fs_default.txt file from the ml mrtrix3 3.0.3 singularity container to the subject’s folder

cp /opt/mrtrix3-3.0.3/share/mrtrix3/labelconvert/fs-default.txt /data/Test/100307

The command labelconvert will use the parcellation and segmentation output of FreeSurfer to create a new parcellated file in .mif format:

labelconvert aparc+aseg.nii.gz FreeSurferColorLUT.txt fs_default.txt nodes.mif -force

Perform nodes co-registeration:

mrtransform nodes.mif -linear diff2struct_mrtrix.txt -inverse -datatype uint32 nodes_coreg.mif -force

Create a whole-brain connectome which denotes the streamlines between each parcellation pair in the atlas. The “symmetric” option makes the lower and upper diagonal the same, the “scale_invnodevol” option scales the connectome by the inverse of the size of the node:

tck2connectome -symmetric -zero_diagonal -scale_invnodevol -tck_weights_in sift_1M.txt tracks.tck nodes_coreg.mif  nodes.csv -out_assignment  assignments_nodes.csv -force

Viewing the connectome

The generated nodes.csv file can be viewed outside neurodesk as a matrix in Matlab.

imagesc(connectome,[0 1])

Structural_connectivity –>

Last modified April 11, 2022 : Minor formatting changes (dfcf47d3)