wb-cifti-tutorial

Connectome workbench and GIFTI/CIFTI tutorial

Slides: https://docs.google.com/presentation/d/1aE4J3XDQ8pWLJfAehLS7NBeiLoyq4SzpgZHUf81OkvY/edit?usp=sharing

Hands-on tutorial material below will be updated throughout the demo

Data:

All tutorial data can be found on:

• Khanlab CBS server share @ /srv/khan/shared/data/2021_connectomewb_tutorial

Recommended templates and parcellations to download:

• HCP-MMP
• Schaeffer atlas
• templateflow

Example: myelin mapping

Prerequisites:

• BIDS dataset with T1w and T2w data
• T2w registered and resampled to T1w (e.g. itksnap, ants, greedy)
• fmriprep processed data (including freesurfer)

1. do math on volume space to get a ratio

wb_command -volume-math "clamp(T1w/T2w,0,10)" sub-0001_T1wDividedByT2w.nii.gz -var T1w ../data/bids/sub-0001/anat/sub-0001_acq-MPRAGE_run-01_T1w.nii.gz  -var T2w ../data/bids/sub-0001/anat/sub-0001_acq-SPACE_run-01_desc-regToT1w_T2w.nii.gz 

2. sample this on the midsurface

wb_command -volume-to-surface-mapping sub-0001_T1wDividedByT2w.nii.gz ../data/fmriprep_20.2.1/sub-0001/anat/sub-0001_acq-MPRAGE_run-1_hemi-L_midthickness.surf.gii sub-0001_hemi-L_myelin.func.gii -trilinear

wb_command -volume-to-surface-mapping sub-0001_T1wDividedByT2w.nii.gz ../data/fmriprep_20.2.1/sub-0001/anat/sub-0001_acq-MPRAGE_run-1_hemi-R_midthickness.surf.gii sub-0001_hemi-R_myelin.func.gii -trilinear

3. visualize in wb_view

4. some other things you could do here:

optional: get clusters of high myelin

wb_command -metric-find-clusters sub-0001_acq-MPRAGE_run-1_hemi-L_midthickness.surf.gii sub-0001_hemi-L_smooth-3mm_myelin.func.gii 2.5 50 sub-0001_hemi-L_smooth-3mm_myelinclusters.func.gii

optional: calculate a gradient

wb_command -metric-gradient ?h.midthickness.surf.gii ?h.myelin.func.gii ?h.myelingradient.func.gii

Set range for color mapping to metric files

wb_command -metric-palette ?h.myelin.func.gii MODE_USER_SCALE -pos-user 0 0.05 -interpolate true -palette-name videen_style -disp-pos true -disp-neg false -disp-zero false

Templates

Now, what if we want to use an atlas parcellation or visualize on a standard template??

1. We need to register between a subject and a template. Thankfully Freesurfer does this already (other common method is MSM - multimodal surface mapping).

The registration output is the vertices of the subject surface, moved onto a sphere. This is surf/lh.sphere.reg in Freesurfer's output. This is registered to a surface template called fsaverage. The other common template is fsLR.

We can first convert this to gifti using

mris_convert data/freesurfer_7.1/sub-0001/surf/lh.sphere.reg myelin_mapping/sub-0001_acq-MPRAGE_run-1_hemi-L_sphere.surf.gii

Now, this has a surface geometry (# of vertices, or density) that is specific to that subject, and is different for each subject. But since it is aligned in the sphere, we can use it to resample things (surface, metrics, labels) to another geometry.

1. resampling subject surfaces to fsLR

This resamples the subject midthickness surface to the fsLR geometry (note: here the fsLR is in fsaverage space - turns out this is just a simple rotation of the sphere, but we don't have to worry about it, we just need to make sure we use the fsLR sphere that is in space-fsaverage)

wb_command -surface-resample sub-0001_acq-MPRAGE_run-1_hemi-R_midthickness.surf.gii  sub-0001_acq-MPRAGE_run-1_hemi-R_sphere.surf.gii  ../data/templateflow/tpl-fsLR/tpl-fsLR_space-fsaverage_hemi-R_den-32k_sphere.surf.gii BARYCENTRIC sub-0001_acq-MPRAGE_run-1_hemi-R_space-fsLR_den-32k_midthickness.surf.gii

wb_command -surface-resample sub-0001_acq-MPRAGE_run-1_hemi-L_midthickness.surf.gii  sub-0001_acq-MPRAGE_run-1_hemi-L_sphere.surf.gii  ../data/templateflow/tpl-fsLR/tpl-fsLR_space-fsaverage_hemi-L_den-32k_sphere.surf.gii BARYCENTRIC sub-0001_acq-MPRAGE_run-1_hemi-L_space-fsLR_den-32k_midthickness.surf.gii

Now you can try visualizing standard 32k data on the subject in wb_view.

2. resampling subject metrics

wb_command -metric-resample sub-0001_hemi-L_smooth-3mm_myelin.func.gii sub-0001_acq-MPRAGE_run-1_hemi-L_sphere.surf.gii ../data/templateflow/tpl-fsLR/tpl-fsLR_space-fsaverage_hemi-L_den-32k_sphere.surf.gii  ADAP_BARY_AREA  sub-0001_hemi-L_smooth-3mm_space-fsLR_den-32k_myelin.func.gii -area-surfs sub-0001_acq-MPRAGE_run-1_hemi-L_midthickness.surf.gii sub-0001_acq-MPRAGE_run-1_hemi-L_space-fsLR_den-32k_midthickness.surf.gii

...

exercise: try mapping an atlas parcellation back to a subject's T1w volume

CIFTI

In addition to the surface-based processing above, the combined surf+vol ("Grayordinates") are another reason to work with CIFTI.

The dense domain is made up of l/r cortical surfaces, and subcortical/cerebellar GM regions