Generate Neural Input Function from Event Timing

Converts event timing information into a neural input function representing the underlying neural activity before HRF convolution. This function is useful for:

Usage

neural_input(x, ...)

# S3 method for class 'Reg'
neural_input(x, start = 0, end = NULL, resolution = 0.33, ...)

Arguments

x: A regressor object containing event timing information
...: Additional arguments passed to methods
start: Numeric; start time of the input function
end: Numeric; end time of the input function
resolution: Numeric; temporal resolution in seconds (default: 0.33)

Value

A list containing:

time: Numeric vector of time points
neural_input: Numeric vector of the stimulus density at each time point (see method details)

Details

stimulus: Creating stimulus functions for fMRI analysis
modeling: Modeling sustained vs. transient neural activity
inputs: Generating inputs for HRF convolution
visualization: Visualizing the temporal structure of experimental designs

For a Reg object the input is built exactly as the default "conv" engine of evaluate builds it, on the time grid seq(start, end, by = resolution). A point event (duration 0) is an impulse of mass amplitude, split linearly between the two samples bracketing its onset. A block is projected onto linear (hat) basis functions centred on the samples, with both edges at their exact positions, so it carries mass amplitude * duration (divided by duration when the regressor has summate = FALSE), however short it is. The values are a density, mass per second: a block is amplitude high away from its edges, a point event is a spike of area amplitude, and sum(neural_input) * resolution is the input mass inside the window. Convolving neural_input * resolution with the HRF sampled every resolution seconds is the convolution that evaluate(x, grid, precision = resolution) computes.

Examples

# Create a regressor with multiple events
reg <- regressor(
  onsets = c(10, 30, 50),
  duration = c(2, 2, 2),
  amplitude = c(1, 1.5, 0.8),
  hrf = HRF_SPMG1
)

# Generate neural input function
input <- neural_input(reg, start = 0, end = 60, resolution = 0.5)

# Plot the neural input function
plot(input$time, input$neural_input, type = "l",
     xlab = "Time (s)", ylab = "Neural Input",
     main = "Neural Input Function")


# Create regressor with varying durations
reg_sustained <- regressor(
  onsets = c(10, 30),
  duration = c(5, 10),  # sustained activity
  amplitude = c(1, 1),
  hrf = HRF_SPMG1
)

# Generate and compare neural inputs
input_sustained <- neural_input(
  reg_sustained,
  start = 0,
  end = 60,
  resolution = 0.5
)