Warning

This repo has moved to https://gitlab.com/flywheel-io/flywheel-apps/bids-hcp

flywheel/hcp-func

Flywheel Gear that runs the functional preprocessing steps of the Human Connectome Project Minimal Preprocessing Pipeline (MPP) described in Glasser et al. 2013. Currently, this includes v4.0-alpha release of fMRIVolume and fMRISurface, as well as generating some helpful QC images. For more info on the pipelines, see HCP Pipelines.

Important notes

• All MRI inputs (fMRI time series, FieldMaps) must include BIDS-conformed DICOM metadata!
• Gradient nonlinearity correction (using coefficient file) is currently only available for data from Siemens scanners.
• Readout distortion correction using B0 field maps (Field map "Option 1", below) is currently only available for data from Siemens scanners. "TOPUP"-style correction (Field map "Option 2", below) should work for all data (but has not yet been tested).

Required inputs

1. fMRI time series NiFTI
2. Field map for correcting readout distortion
   • Option 1: GRE = "typical" GRE B0 field map including magnitude and phase volumes
   • Option 2: SpinEchoFieldMap = a pair of spin echo with opposite phase-encode directions ("Positive" = R>>L or P>>A, and "Negative" = L>>R or A>>P) for "TOPUP"-style distortion estimation
3. StructZip output from the HCP-Struct gear (containing T1w/, T2w/, and MNINonLinear/ folders)
4. FreeSurfer license.txt file (found in $FREESURFER_HOME/license.txt)

Optional inputs

1. fMRIScout: high-quality exemplar volume from fMRI time-series. If using Multi-Band for fMRI, and Single-Band reference volume is available, use SBRef. Otherwise, leave empty to first time series volume for registration.
2. Gradient nonlinearity coefficients copied from scanner. See FAQ 8. What is gradient nonlinearity correction?
   • If needed, this file can be obtained from the console at C:\MedCom\MriSiteData\GradientCoil\coeff.grad for Siemens scanners
   • Note: This effect is significant for HCP data collected on custom Siemens "ConnectomS" scanner, and for 7T scanners. It is relatively minor for production 3T scanners (Siemens Trio, Prisma, etc.)

Configuration options

1. fMRIName: Output name for preprocessed data (default = rfMRI_REST)
2. BiasCorrection: Bias-field estimation method. 'NONE' (default), 'SEBased', or 'Legacy'. 'SEBased'=Estimate from SpinEchoFieldMap (only possible with both Pos and Neg SpinEcho), 'Legacy'=Estimate from structural scans (only valid if structural collected in the same session, and without any subject movement)
3. MotionCorrection: Use 'MCFLIRT' (standard FSL moco) for most acquisitions. 'FLIRT'=custom algorithm used by HCP internally, but not recommended for public use
4. AnatomyRegDOF: Degrees of freedom for fMRI->Anat registration. 6 (default) = rigid body, when all data is from same scanner. 12 = full affine, recommended for 7T fMRI->3T anatomy
5. RegName: Surface registration to use during CIFTI resampling: either 'FS' (freesurfer) or 'MSMSulc'. ('Empty'=gear uses RegName from HCP-Structural)

Outputs

• <subject>_<fMRIName>_hcpfunc.zip: Zipped output directory containing <fMRIName>/ and MNINonLinear/Results/<fMRIName>/ folders
• <subject>_<fMRIName>_hcpfunc_QC.*.png: QC images for visual inspection of output quality (Distortion correction and registration to anatomy, details to come...)
• Logs (details to come...)

Gear Release Notes

The latest iteration of the hcp gears use a common docker base image to consolidate both library installations and common functionality across gears. See HCP Base Docker Image for details.

Important HCP Pipeline links

• HCP Pipelines
• HCP Pipelines FAQ
• HCP Pipelines v3.4.0 release notes