Identifying cortical areas that underlie the transformation from 2D retinal to 3D head-centric motion signals

Table of Contents

1. Getting Started
   • Download dataset from OpenNeuro
   • Setup GitHub repositories
2. Pipelines
   • dcm2bids
   • fMRIPrep
   • Extract ROIs
3. Analysis
   • TAFKAP
   • Searchlight
4. Visualization

Getting Started

Download dataset from OpenNeuro

• Dataset link: OpenNeuro

• Download with nodejs, S3, or datalad

openneuro download --draft ds004443 ds004443-download/
aws s3 sync --no-sign-request s3://openneuro.org/ds004443 ds004443-download/
datalad install https://github.com/OpenNeuroDatasets/ds004443.git

• Dataset is in BIDS format such that the each subject folder sub-0XXX contains the orginal raw NIfTI files for both anatomical and funcational files, and the folder derivatives/fmriprep contains files that have been processed through fMRIPrep (in T1w space and in fsaverage6 space).
  Read more about the output space here.

Setup GitHub repositories

• Go to Github directory

  cd ~/Documents/GitHub 

• Clone the repositories

  git clone https://github.com/putiw/DecodingPublic.git
  git clone https://github.com/Rokers/TAFKAP.git
  git clone https://github.com/cvnlab/knkutils.git
  git clone https://github.com/cvnlab/cvncode.git
  git clone https://github.com/gllmflndn/gifti.git

(back to top)

Pipelines

dcm2bids

• In this step we converted data from raw DICOM images into NIfTI files and organized data in BIDS format using Dcm2Bids.

• The output of this step is provided on OpenNeuro and is used in the proceeding steps. (The DICOM images will be provided upon request).

fMRIPrep

• We processed raw NIfTI files (those within sub-0XXX folder) using fMRIPrep version: 20.2.6.

Using the fMRIPrep command: /usr/local/miniconda/bin/fmriprep --nprocs 14 --omp-nthreads 14 /scratch/pw1246/MRI/Decoding/rawdata /scratch/pw1246/MRI/Decoding/derivatives participant --fs-license-file /scratch/pw1246/MRI/Decoding/code/license.txt --output-space T1w:res-native fsnative:den-41k MNI152NLin2009cAsym:res-native fsaverage:den-41k --participant_label sub-0201 --skip_bids_validation -w /tmpdata/pw1246/2800748/1 --no-submm-recon

• The output of this step can be found in the derivatives/fmriprep folder.

        * 📝 Note: this step also requires freeSurfer.

Extract ROIs

• We extracted the region of interests (ROIs) using the neuropythy pipeline.

• Example way of running neuropythy with Docker.

docker run -ti --rm -v <path-to-SUBJECTS_DIR>:/subjects nben/neuropythy atlas --verbose <subject-ID> --volume-export

• This step generates the ROI atlas file named wang15_mplbl.mgz in derivatives/freesurfer/sub-xxxx/mri folder.

     * 📝 Note: You can find the output of this step (i.e. pre-generated the ROIs files in appropriate resolution) in derivatives/fmriprep/sub-xxxx/ses-01/anat/rois folder.

     * 📝 Note: Alternatively, you can run extractRois.m in the current repo to generate the same files.

     * 📝 Note: A better way: atlasmgz (the paper did not use this way)

Analysis

TAFKAP

• This analysis resulted in figure 3 to 6.
  See orginal paper here.
  See the code for this analysis here: run_TAFKAP.

• The input to this code is the fmriprep processed NIfTI files in derivatives/fmriprep/sub-xxxx/ses-xx/func folder.
  The file name should contain "*space-T1w_desc-preproc_bold.nii.gz".

• The only variable that needs user input is the bidsDir, which should be set to whichever the directory of the project is in.

• The code will print out the decoding accuracy for each ROI and for each subject as the output in the command window.

• You can play with different ways of denoise, detrend, and estimate response amplitude in the sub-function load_vol.m

Searchlight

• This analysis resulted in figure 7 and 8.
  See the code for this analysis here: run_Searchlight.

• The required input files to the searchlight analysis are the functional scans in fsaverage6 space in the derivatives/fmriprep/sub-xxxx/func/ folder.
  The file name should contain "*space-fsaverage6_hemi-R_bold.func.gii".

• You will also need the inflate surface files in the derivatives/freesurfer/fsaverage6/mri folder.

• The output is the 81924 by 1 matrix containing one decoding accuracy value for each vertex in the fsaverage6 space.
  This result can be then used in further analysis.

• You will need to manually define three directories.

bidsDir = '/Volumes/Vision/MRI/DecodingPublic'; % project directory containing the folder **derivatives**.
fsDir = '/Applications/freesurfer/7.2.0'; % freesurfer directory, most likely in application folder.
gitDir = '~/Documents/GitHub'; % GitHub directory, where you installed all the repositories. 
set_up(bidsDir,gitDir,fsDir)

• You can play with different ways of denoise, detrend, and estimate response amplitude in the sub-function load_surf.m

Visualization

• See plot_figures.m

Contact

Puti Wen - pw1246@nyu.edu

(back to top)