Build a streaming loader backed by neuroim2 objects

The returned loader exposes `n()`, `X(s)`, `B(s)`, and `Omega(s)` methods compatible with streaming DKGE fits.

Usage

dkge_neuro_loader(design_objs, bv_list, labels_list = NULL, omega_fun = NULL)

Arguments

design_objs

List of design matrices or `fmridesign` objects.

bv_list

List of `neuroim2::NeuroVec` objects, `ClusteredNeuroVec` objects, or file paths readable by `neuroim2::read_vec()`.

labels_list

List of `neuroim2::NeuroVol` label volumes (same length as `bv_list`). Can be NULL if bv_list contains ClusteredNeuroVec objects.

omega_fun

Optional function mapping a label volume or ClusteredNeuroVec to cluster weights. Default uses cluster sizes as weights.

Value

Loader list suitable for streaming DKGE fits.

Examples

# \donttest{
if (requireNamespace("neuroim2", quietly = TRUE)) {
  labels <- neuroim2::read_vol(system.file("extdata", "global_mask2.nii.gz", package = "neuroim2"))
  vols <- lapply(1:6, function(i) labels * i)
  bv <- neuroim2::vec_from_vols(vols)
  x_mat <- cbind(intercept = 1, trend = seq_len(length(vols)))
  loader <- dkge_neuro_loader(
    design_objs = list(x_mat, x_mat),
    bv_list = list(bv, bv),
    labels_list = list(labels, labels)
  )
  loader$n()
}
#> [1] 2
# }