fNIRS App: GLM Pipeline

This fNIRS App will runs a GLM pipeline on your data and export subject and group level results.

Feedback is welcome!! Please let me know your experience with this app by raising an issue.

Usage

To run the app you must have docker installed. See here for details about installing fNIRS Apps. You do NOT need to have MATLAB or python installed, and you do not need any scripts. See this tutorial for an introduction to fNIRS Apps.

To run the app you must inform the software where the bids_dataset resides. This is done by passing the location of the dataset using the -v command to the app. To run this app use the command:

docker run -v /path/to/data/:/bids_dataset ghcr.io/rob-luke/fnirs-apps-glm-pipeline/app

By default the app will process all subject and tasks. You can also specify additional parameters by passing arguments to the app. A complete list of arguments is provided below. A more complete example that only runs on participant 3 and also specifies that short channel regression should not be used, can be run as:

docker run -v /path/to/data/:/bids_dataset ghcr.io/rob-luke/fnirs-apps-glm-pipeline/app \
    --subject-label 03 \
    --short-regression False

Pipeline details

The pipeline converts the data to optical density.
Then applies the Beer Lambert Law conversion.
The data is resampled
- to 0.6 Hz. Use --sample-rate X to resample to user specified value.
A GLM is applied
- The model used is a glover HRF convolved with boxcar of stimulus duration
- If --short-regression True is specified the short channels will be added as regressors to the design matrix for the GLM computation.
- Drift components will be added to the design matrix using a cosine model including frequencies up to 0.01 Hz (TODO: make user specified parameter).
The results of the GLM will be exported per subject as a tidy csv file per run.
A mixed effects model is then run on the individual data to produce a summary result
- First, for each subject the channels are combined to a single region of interest
- The model is then applied that includes condition and chromaphore as a factor with id as a random variable
- A summary is exported as a text file

Arguments

	Required	Default	Note
short-regression	optional	True	Include short channels as regressor.
export-drifts	optional	False	Export the drift coefficients in csv.
sample-rate.	optional	0.6.	Sample rate the data will be resampled to.
export-shorts	optional	False	Export the short channel coefficients in csv.
subject-label	optional	[]	Sujects to process. Default is to process all.
task-label	optional	[]	Tasks to process. Default is to process all.

For example

docker run -v /path/to/data/:/bids_dataset ghcr.io/rob-luke/fnirs-apps-glm-pipeline/app --short-regression True --export-shorts True --subject-label 02 04

Updating

To update to the latest version run.

docker pull ghcr.io/rob-luke/fnirs-apps-glm-pipeline/app

Or to run a specific version:

docker run -v /path/:/bids_dataset ghcr.io/rob-luke/fnirs-apps-glm-pipeline/app:v1.4.2

Additional information

Boutiques

This app is boutiques compatible. In addition to the methods described above, this app can also be run using boutiques bosh command. You can see an example usage of this app with boutiques at https://github.com/rob-luke/fnirs-apps-demo.