Skip to contents

Cluster4d wrapper for ReNA++

Source: R/rena_plus.R
cluster4d_rena_plus.Rd

Cluster4d wrapper for ReNA++

Usage

cluster4d_rena_plus(
  vec,
  mask,
  n_clusters = 100,
  spatial_weight = 0.5,
  r = 2,
  lambda = 1,
  grad_img = NULL,
  connectivity = 26,
  max_iterations = 50,
  verbose = FALSE,
  ...
)

Arguments

vec

A NeuroVec instance supplying the 4D data (x, y, z, time) to cluster

mask

A NeuroVol mask defining the voxels to include in clustering. Values must be finite. Voxels are included exactly when the mask value is strictly greater than zero; zero and negative values are excluded.

n_clusters

Exact finite integer target number of clusters, from one through the number of included mask voxels.

spatial_weight

Method-specific spatial control in [0, 1]. For methods that define a convex feature/spatial blend, zero disables the spatial term and one disables the feature term. Other methods use the value as a bounded mapping to a native scale; their endpoints do not have that blend meaning. Default 0.5. This parameter is inactive for "rena" and "rena_plus"; supplying it explicitly for either method is an error. Maps to method-specific parameters:

  • supervoxels: alpha = 1 - spatial_weight (0 = all spatial, 1 = all feature)

  • snic: compactness = spatial_weight * 10 (range 0-10)

  • slic: compactness = spatial_weight * 20 (typical range 1-20)

  • corr_slic/brs_slic: direct convex blend between correlation and scaled spatial distance

  • slice_msf: compactness = spatial_weight * 10, followed by FH component scale 2 / (compactness + 1). This changes the graph-segmentation scale; it is not a convex spatial/feature blend.

  • flash3d: lambda_s = spatial_weight (direct mapping)

  • mcl: direct blend between feature and spatial edge similarities

r

Over-clustering factor for coarsening stage.

lambda

Gradient penalty weight.

grad_img

Optional gradient/intensity vector (see rena_plus).

connectivity

Integer neighborhood connectivity. Supported values are method-specific: supervoxels accepts 6, 18, 26, or 27; slic, corr_slic, brs_slic, and slice_msf accept 6 or 26; G3S, ReNA variants, and MCL accept 6, 18, or 26. It is inactive for snic, flash3d, acsc, and commute.

max_iterations

Positive finite integer iteration limit for iterative methods. It is inactive for "snic", "slice_msf", and "commute"; supplying it explicitly for those methods is an error.

verbose

Print progress messages. Default FALSE.

...

Additional method-specific parameters. See method documentation for details.

Value

A cluster4d_result.