Extending morphological pattern segmentation to 3D voxels

This short communication introduces the logic, demonstrates its use, and identifies the availability of a new tool that extends the traditional 2D morphological segmentation of binary raster data into the 3-dimensional realm of voxels. A combination of 3-dimensional array data and network graph theory are implemented to facilitate the logical parsing of identified 3-dimensional features into their mutually exclusive constituent morphological classes. All processing is performed in the R environment, providing the ability for anyone to perform the demonstrated analyses on their own data. The only input requirement is a binary (1 = feature of interest, 0 otherwise) 3-dimensional array, where each voxel of interest is then classified into classes called outside, mass, skin, crumb, antenna, circuit, bond, and void that correspond their 2-dimensional equivalents of background, core, edge, islet, branch, loop, bridge, and perforation. An additional class called the void-volume identifies voxels belonging to the empty space within the object of interest. The work helps to bring pattern metrics into the 3-dimensional world, particularly given the reliance on adjacency and connectivity assessments