Skip to contents

A parcel_space owns only the feature-axis algebra required by fmridataset: stable parcel identity, a parent feature space, and explicit parent-to-parcel operators. Atlas discovery, labels, and provenance remain owned by neuroatlas and can be imported with parcel_space_from_atlas().

Usage

parcel_space(
  parent,
  parcel_ids,
  membership,
  data = NULL,
  atlas,
  aggregation = c("mean", "sum"),
  decoder = NULL,
  metadata = list()
)

Arguments

parent

Parent feature_space (usually a volume or surface space).

parcel_ids

Stable atlas-native parcel identifiers.

membership

Non-negative parent-feature by parcel membership weights.

data

One metadata row per parcel. The id, label, and hemi conventions match neuroatlas::parcel_data when present.

atlas

Atlas identity list containing at least id, or a scalar ID.

aggregation

Either weighted "mean" or weighted "sum".

decoder

Optional parent-feature by parcel reconstruction operator. The default blends overlapping parcel values by row-normalized membership.

metadata

Additional serializable metadata.

Value

A parcel_space.

Examples

parent <- volume_space(c(3, 2, 1), support = 1:6)
membership <- Matrix::sparseMatrix(
  i = 1:6, j = c(1, 1, 1, 2, 2, 2), x = 1, dims = c(6L, 2L)
)
x <- parcel_space(parent, c(10L, 20L), membership, atlas = "toy-atlas")
n_features(x)
#> [1] 2