Matlab Assistant for Surface-based fMRI Analysis

DOI: 10.17605/OSF.IO/D76FM

(Working in progress...)
Last updated: 2023

• Matlab Assistant for Surface-based fMRI Analysis
• Introduction
  • Dependency
• Preparations
  • Setup global environment
  • Create files needed for FreeSurfer
• Pre-processing
  • FreeSurfer
• First-level analysis
  • Configure analysis, contrasts and perform the analysis
  • Draw labels based on contrast
• Group level analysis
  • 1. Group level analysis on fsaverage
  • 2. Univariate analysis of ROIs on self surface
  • 3. Classification/decoding with CoSMoMVPA
  • 4. Searchlight with CoSMoMVPA
  • Some suggestion for these analyses (or notes for myself)
• Data visualization
  • Quality assurance of recon-all results
  • Check other results
• How-to
• Q&A

Introduction

These Matlab functions mainly call FreeSurfer commands to analyze fMRI data on the surface, perform multivariate pattern analysis (MVPA) with CoSMoMVPA toolbox, and visualize results with FreeView and codes based on Dr. Kendrick Kay's work.

Now some wrappers for BIDS, fMRIPrep, and HCPPipeline are also included.

Note:

1. These functions were only tested in Mac.
2. These functions were built based on FreeSurfer 6.0, and therefore some may fail when earlier FreeSurfer versions are loaded (it seems to work on later versions).
3. Some default parameters in the functions were set according to our data acquisition protocol, they may not be appropriate for others.

Dependency

The following software and toolboxes should be installed properly before using this toolbox. Also, it is assumed that the user understands the FS-FAST directory structure and the general steps (and commands) for performing fMRI data analysis in FS-Fast.

• FreeSurfer;
• NIfTI_20140122;
• CoSMoMVPA (for running multivariate pattern analysis);
• knkutils and cvncode (for visualizations in Matlab).

Check here (to be added) for a full list of dependended toolboxs.

Preparations

Setup global environment

As most of the functions in this toolbox call FreeSurfer commands which are linux commands, the path to FreeSurfer needs to be added to the global environment $PATH. In addition, the matlab/ folder in FreeSurfer ($FREESURFER_HOME/matlab/) also needs to be added to Matlab path. These can be set up in two ways.

• Method 1:

  1. Start a new terminal;
  2. Set all the necessary global environment variables (e.g., $FREESURFER_HOME, $SUBJECTS_DIR and $FUNCTIONALS_DIR) and set up FreeSurfer (the instruction is available on the FreeSurfer website);
  3. Start Matlab in the same terminal by running path-to-matlab-folder/bin/matlab in the terminal. [You may also add the path-to-matlab-folder/bin to $PATH by following this instruction and run matlab in the terminal].

• Method 2:

  1. Start Matlab and add all subdirectory of this toolbox to the Matlab path;
  2. Run fs_setup('path/to/freesurfer/home'); to set 'path/to/freesurfer/home' as $FREESURFER_HOME and add necessary paths to $PATH;
  3. Run fs_subjdir('path/to/subjects/dir'); to set 'path/to/subjects/dir' as $SUBJECTS_DIR;

Create files needed for FreeSurfer

Make sure the following files have been created properly (at least before running the general linear model (GLM) in FreeSurfer).

• create paradigm files (parfiles);
• run files (if needed): txt files containing the name of runs (folders) in a column. For example, you should have different run files for localizer and main runs.
• subjectname: a txt file containing the subject name in $SUBJECTS_DIR; this links the functional data folder and the structural data folder for that subject (more see here).
• sessID Files: a txt file containing the list of session names in that folder (more see here).

Pre-processing

Data can be pre-processed by FreeSurfer, fMRIPrep or HCPpipeline. After pre-processing, there should be one folder consisting the functional data and a separate folder consisting structural data (i.e., the recon-all results).

FreeSurfer

In terminal:

# structure
recon-all -s subjCode -i path/to/T1/image -T2 path/to/T2/image -all 

# functional
preproc-sess -sf sessFile -fsd bold -surface self lhrh -mni305 -fwhm 0 -per-run -force 

In Matlab:

% structure
fs_recon()

% functional
fs_preproc()

First-level analysis

Configure analysis, contrasts and perform the analysis

Steps	functions
1.Configure analyses	fs_mkanalysis()
2.Configure contrasts	fs_mkcontrast()
3.Perform the analysis	fs_selxavg3()
More information on configuring analyses and contrasts can be found here.

Draw labels based on contrast

After performing the first level analysis, you may want to create ROI label files based on the contrast. Please check How to draw ROI on surface?

Group level analysis

1. Group level analysis on fsaverage

More information can be found here.

Steps	functions
1.Concatenate first level results	fs_isxconcat()
2.Perform group level GLM	fs_glmfit_osgm()
3.Multiple comparison correction with permutation	fs_glmfit_perm()

2. Univariate analysis of ROIs on self surface

To perform statistical analyses on self surface for different labels (ROIs) later, use fs_cosmo_readdata() to load the surface data as a table in Matlab.

3. Classification/decoding with CoSMoMVPA

To perform N-fold cross-validation classification/decoding, use fs_cosmo_cvdecode().

4. Searchlight with CoSMoMVPA

To perform N-fold cross-validation searchlight on the surface, use fs_cosmo_sesssl(). (Multiple comparison corrections with TFCE will be added later.)

Some suggestion for these analyses (or notes for myself)

Analysis	Template	Smooth	Runwise
1.Group level analysis	fsaverage	sm5	no
2.ROI (label) univariate	self	sm5	no
3.ROI (label) decoding	self	sm0	yes
4.Searchlight	fsaverage	sm0	yes

Data visualization

Visualize nifty files: ITK-SNAP. Visualize surface data: FreeView or functions based on knkutils and cvncode.

Quality assurance of recon-all results

To check the recon-all results, use fv_checkrecon('subjcode'). To check both the volumetric and surface data, use fv_checkreg(). To check the volumetric data only, run fv_vol('', subjCode) . To check the surface data only, run fv_surf('', 'inflated') and then select which participants to be displayed in FreeView.

Check other results

Results for visualization	functions
first level results (i.e., the screenshots of contrasts)	fs_cvn_print1st()
second level results	fs_cvn_print2nd()
overlapping between labels	fs_labeloverlap()
labels files	fs_cvn_print1st()

How-to

• How to set up FreeSurfer in the terminal?
• How to draw ROI on surface?
• How to trim a label file with some restrictions (e.g., 100 vertices)?
• How to visualize label files (i.e., ROIs)?
• How to visualize activation maps?

Q&A

• How to run linux commands via Matlab?
  For most cases, you can use the Matlab function system() to run linux commands (e.g., system('linux commands')). However, system() does not always do what you want to do. For example, for setting the global environment variable (e.g., $FREESURFER_HOME), the Matlab function setenv() has to be used (e.g., setenv('FREESURFER_HOME', '/Applications/freesurfer/')).

• Question
  The answers.