[WIP] GLM

GLM design and fitting

The process NIRS - wip -> GLM - design and fit takes care of building the design matrix from the stimulation events and fits the GLM on the input data.

Inputs

It can handle data in the channel-space or projected on the cortical surface (TODO: see).

Process parameters

• Stimulation events: list of event group names for which to build stimulus-induced regressors.
• HRF model: model to build the hemodynamic time course. Note that the HRF duration is fixed to 25 sec.
  • CANONICAL: canonical HRF as in Glover 1995
  • GAMMA: sum of two gamma functions
  • BOXCAR: binary boxcar / step function
• Add constant regressor: append a column of ones to the design matrix.
• Fitting method:
  • OLS: ordinary least square (white noise).
  • AR-ILS: autoregressive noise model. This method uses the Analyzir toolbox which must be installed prior to running this process.
• Extra outputs: export intermediary results.

Outputs

A custom brainstorm object containing all GLM results is produced, named as <input data> | GLM <fit method> - results. It exposes the design matrix, which can be viewed by double clicking on it.

Extra outputs (if enabled in parameters):

• Residuals: residual time-series after model fitting ( Y - X * B_fit )
• Beta maps: maps of estimated beta.

Intra subject contrast and stat maps

See process NIRS - wip -> GLM - intra subject contrast.

Input

The contrast computation takes the results of the previous process as input (<input data> | GLM <fit method> - results)

Process parameters