Skip to contents

First-level fNIRS cortical-activation GLM

Source: R/activation-glm.R
nirsActivationGLM.Rd

Fits a general linear model of every haemoglobin channel against HRF-convolved task regressors: for each condition, a boxcar over its stimulus blocks convolved with a canonical double-gamma HRF, plus a discrete-cosine drift basis and optional short-separation nuisance regressors. Returns the per-channel activation amplitude (beta), standard error and t-statistic for each condition. Run it separately on the HbO and HbR assays.

Usage

nirsActivationGLM(
  x,
  events = NULL,
  assay_name = "HbO",
  conditions = NULL,
  basis = c("canonical", "derivative"),
  drift = 6,
  nuisance = NULL,
  hrf = list(peak = 6, undershoot = 16, ratio = 6),
  prewhiten = c("none", "ar1")
)

Arguments

x

A PhysioExperiment of haemoglobin concentration (e.g. the HbO assay from mbll()).

events

A data.frame with numeric onset and duration (seconds) and a condition column; if NULL, the object's events (PhysioExperiment::getEvents onset/duration/value) are used, with value as the condition.

assay_name

Assay to model (default "HbO").

conditions

Optional character vector selecting/ordering conditions.

basis

"canonical" (HRF only) or "derivative" (HRF + temporal derivative per condition).

drift

Number of discrete-cosine drift terms (default 6).

nuisance

Optional numeric matrix of extra nuisance regressors (e.g. shortSeparationDesign() output).

hrf

List of HRF parameters (peak, undershoot, ratio).

prewhiten

"none" or "ar1" (per-channel AR(1) prewhitening; strongly recommended for fNIRS to avoid inflated t-statistics).

Value

An S3 nirs_activation_glm with coefficients (a channel x condition data.frame of beta, se, t, df, p), the beta/se matrices, per-channel sigma2/rho, and the design (for nirsActivationContrast()).

References

Friston KJ et al. (1994) Statistical parametric maps. Hum Brain Mapp 2:189-210. Huppert TJ et al. (2009) HomER. Appl Opt 48:D280.

See also

nirsActivationContrast(), shortSeparationDesign(), mbll()

Examples

hb <- mbll(intensityToOD(PhysioNIRS:::.nirs_demo_object()))
fit <- nirsActivationGLM(hb, assay_name = "HbO")
fit$coefficients
#>   channel condition       beta        se         t  df         p
#> 1   S1_D1      task 0.03532496 0.1592257 0.2218547 152 0.8247248
#> 2   S2_D2      task 0.13246861 0.5970962 0.2218547 152 0.8247248