Expands the boundaries of brain atlas parcels by dilating them into adjacent unassigned voxels within a specified mask. This is useful for filling small gaps between parcels or extending parcels into neighboring regions.
Arguments
- atlas
An object of class "atlas" containing the parcellation to be dilated
- mask
A binary mask (NeuroVol object) specifying valid voxels for dilation. Dilation will only occur within non-zero mask values. Can also be a character string representing a TemplateFlow space ID (e.g., "MNI152NLin2009cAsym"), in which case the corresponding standard brain mask will be fetched from TemplateFlow. Alternatively, can be a list of arguments to pass to `neuroatlas::get_template` to fetch a specific mask.
- radius
Numeric. The maximum distance (in voxels) to search for neighboring parcels when dilating. Default: 4
- maxn
Integer. Maximum number of neighboring voxels to consider when determining parcel assignment. Default: 50
Value
A ClusteredNeuroVol object containing the dilated parcellation.
The object maintains the original label mappings but may include additional
voxels in existing parcels.
Details
The dilation process:
Identifies unassigned voxels within the mask that are adjacent to existing parcels
For each unassigned voxel, finds nearby assigned voxels within the specified radius
Assigns the unassigned voxel to the nearest parcel
Respects mask boundaries to prevent dilation into unwanted regions
The function uses a k-d tree implementation (via rflann) for efficient nearest neighbor searches in 3D space.
References
The algorithm uses the FLANN library for efficient nearest neighbor searches: Muja, M., & Lowe, D. G. (2014). Scalable nearest neighbor algorithms for high dimensional data. IEEE Transactions on Pattern Analysis and Machine Intelligence, 36(11), 2227-2240.
See also
get_template_brainmask for creating appropriate masks from TemplateFlow
Examples
if (FALSE) { # \dontrun{
# Load an atlas
atlas <- get_aseg_atlas()
# Use TemplateFlow brain mask
dilated <- dilate_atlas(atlas, "MNI152NLin2009cAsym", radius = 4)
# More conservative dilation with fewer neighbors
dilated_conservative <- dilate_atlas(atlas, "MNI152NLin2009cAsym", radius = 2, maxn = 20)
} # }