Page tree

Versions Compared


  • This line was added.
  • This line was removed.
  • Formatting was changed.


10. What is the order of pipelines for resting-state data?

Adapted from a 6/30/2017 post to the HCP-Users list by Matt Glasser:

If you wish to use MSMSulc and MSMAll registration (in conjunction with a recent version of the HCP Pipelines), the MSM binary is available here:

i) PreFreeSurfer
ii) FreeSurfer
iii) PostFreeSurfer (using MSMSulc registration)
iv) fMRIVolume
v) fMRISurface
vi) ICA+FIX (hcp_fix or hcp_fix_multi_run), followed by PostFix if you want to review the classifications
vii) MSMAll (Do MSM Surface Registration)
viii) GroupDeDrift (Compute and Group drift, can skip if you don’t want to)
ix) DedriftAndResample (Apply MSMAll)
xDeDriftAndResample (Apply the MSMAll registration)  (MSMRemoveGroupDrift is for the uncommon case of generating a new registration template).
ix) *Global Noise Removal* (i.e., temporal ICA; not available publicly yet, but will handle residual global noise in the data without removing the global signal)
xix) Resting state analysis

For a resting state analysis, you could do something like dual regression and 'fslnets' to make netmats.  You could also use the HCP’s multi-modal parcellation to parcellate the MSMAll aligned dense timeseries into a parcellated timeseries, and then use fslnets to estimate functional connectivity (can be either full correlation or partial correlation—which is better if step 10 has not been done).  You can also use the MIGP algorithm to combine resting state fMRI data across subjects for group ICA or group dense functional connectivity.  You can use wb_command -cifti-correlation to compute functional connectivity as well.