deconvolve_nlreg_r: R port of Bush and Cisler 2013, Magnetic Resonance Imaging...

In PennStateDEPENdLab/dependlab: Useful functions for the UNC Chapel Hill DEPENd Lab (Director: Michael Hallquist)

R port of Bush and Cisler 2013, Magnetic Resonance Imaging Adapted from the original provided by Keith Bush

Description

R port of Bush and Cisler 2013, Magnetic Resonance Imaging Adapted from the original provided by Keith Bush

Usage

deconvolve_nlreg_r(
  BOLDobs,
  kernel,
  nev_lr = 0.01,
  epsilon = 0.005,
  beta = 40,
  normalize = TRUE,
  trim_kernel = TRUE
)

Arguments

BOLDobs	observed BOLD timeseries
kernel	assumed kernel of the BOLD signal (e.g., from spm_hrf)
nev_lr	learning rate for the assignment of neural events. Default: .01
epsilon	relative error change (termination condition). Default: .005
beta	slope of the sigmoid transfer function (higher = more nonlinear)
normalize	whether to unit-normalize (z-score) BOLDobs before deconvolution. Default: TRUE
trim_kernel	whether to remove the first K time points from the deconvolved vector, corresponding to kernel leftovers from convolution. Default: TRUE

Details

This function deconvolves the BOLD signal using Bush 2011 method

Author: Keith Bush, PhD Institution: University of Arkansas at Little Rock Date: Aug. 9, 2013

The original code did not unit normalize the BOLD signal in advance, but in my testing, this proves useful in many cases (unless you want to mess with the learning rate a lot), especially when the time series has a non-zero mean (e.g., mean 100).

Value

A time series of the same length containing reconstructed neural events

Author(s)

Keith Bush

PennStateDEPENdLab/dependlab documentation built on Sept. 26, 2024, 8:40 p.m.