Skip to contents

fmrihrf

Hemodynamic response functions and event-related regressors for fMRI analysis in R.

fmrihrf provides a composable toolkit for constructing, modifying, and convolving HRFs with experimental designs. It ships with standard basis sets (SPM canonical, B-spline, Fourier, FIR, Gamma, Gaussian, and more), decorators for time-shifting and blocking, and a fast C++ convolution backend.

Installation 

# From CRAN
install.packages("fmrihrf")

# Development version
remotes::install_github("bbuchsbaum/fmrihrf")

Quick start 

library(fmrihrf)

# Evaluate the SPM canonical HRF over 0-30 seconds
t <- seq(0, 30, by = 0.1)
y <- evaluate(HRF_SPMG1, t)
plot(t, y, type = "l", xlab = "Time (s)", ylab = "Response")

# Build a regressor from event onsets
reg <- regressor(onsets = c(2, 10, 18), hrf = HRF_SPMG1,
                 duration = 0, amplitude = 1,
                 span = 24, sampling_frame = sampling_frame(blocklens = 100, TR = 1))
plot(reg)

Key features

Multiple basis sets — Use a single canonical HRF or a flexible basis set to capture response variability.

HRF_SPMG1                                    # SPM canonical (double gamma)
HRF_SPMG3                                    # canonical + temporal & dispersion derivatives
hrf_bspline(t, N = 6)                        # B-spline basis
hrf_fourier(t, N = 5)                        # Fourier basis

Decorators — Modify any HRF through functional composition.

lag_hrf(HRF_SPMG1, lag = 2)                  # shift peak by 2 s
block_hrf(HRF_SPMG1, width = 15)             # sustained/blocked response
normalise_hrf(HRF_SPMG1)                     # unit peak-normalised

Custom HRFs — Wrap any f(t) into the HRF system.

my_hrf <- as_hrf(function(t) exp(-t / 5), name = "exponential", span = 20)
evaluate(my_hrf, seq(0, 20, by = 1))

Regressor construction — Convolve events with HRFs to produce design-matrix columns, with support for variable durations, amplitudes, and multi-basis expansion.

sf <- sampling_frame(blocklens = c(200, 200), TR = 2)
reg <- regressor(onsets = c(10, 30, 50), hrf = HRF_SPMG1,
                 duration = c(0, 5, 0), amplitude = c(1, 1.5, 1),
                 sampling_frame = sf)
evaluate(reg)

Fast convolution — Core routines are implemented in C++ (Rcpp / RcppArmadillo) for efficient large-scale design matrix generation.

Documentation

Full documentation is available at https://bbuchsbaum.github.io/fmrihrf/.

Vignettes cover the main workflows:

License

MIT