label is an atlas's join key: core, the palette's names and every
geometry payload (geom, vertices, meshes, centerlines) are keyed on
it. Rewriting it in one of those places and not the others silently decouples
geometry from metadata, which is how a renamed region ends up drawn in the
wrong colour or not at all, so this rewrites all of them together.
Value
The ggseg_atlas with mapping applied to core$label, the
palette's names and every payload slot that carries a label column.
Details
Labels absent from mapping are left alone, so contextual geometry such as
the cortex_ silhouette survives a partial re-key untouched.
This is the supported way to correct an atlas's identifiers – the operation
tracula() itself needed when its .bbr.prep suffix was restored. Use
atlas_region_rename() instead to change the display names in region,
which is not a join key.
See also
atlas_region_rename() to rename regions rather than re-key labels.
Other atlas manipulations:
atlas_manipulation,
atlas_structure_reorder()
Examples
a <- relabel_atlas(aseg(), c("Left-Thalamus" = "lh_thalamus"))
head(atlas_labels(a))
#> [1] "Left-Cerebellum-Cortex" "lh_thalamus" "Left-Caudate"
#> [4] "Left-Putamen" "Left-Pallidum" "Brain-Stem"