Cleanse Structure removes small, connected components from the structure. It is well suited to filter out the noise in a structure generated after importing and thresholding a 3D grey-value image from a CT-scan.

The structure can be cleansed with periodic boundary conditions in all directions, or only in the selected direction(s) (Periodicity in X-Direction, Y-Direction and/or Z-Direction). If no direction is chosen it is cleansed non-periodically (with symmetric boundaries).

The selections in the Neighborhood panel determine which voxels are detected as connected. Checking Face is more restrictive than choosing Face or Edge. The most permissive Neighborhood condition is Face, Edge or Vertex.

Select the material ID of the connected components to be removed (Material to Cleanse) and the material ID to be applied to the cleansed components (New Material ID). If the material to cleanse is set to All, the New Material ID is the material ID of the largest surrounding connected component.

Cleansed are all connected components of the specified material ID which contain only the given Number of Voxels or less.

In this example, Cleanse Structure is applied to an imported structure with high levels of noise in the CT image, leading to unwanted components. Here, it is shown how small unconnected components, with less than 10 voxels, are removed with the cleansing.

Taking a closer look at the original structure, we see examples for the different cases.

If we choose Face, only voxels connected with face contact to at least 10 other voxels (Number of Voxels option) remain. The red voxel shown above is connected to the other voxels by faces.

If we choose Face or Edge, also voxels connected to at least 10 other voxels over edges remain. The group of green voxels shown above is connected to the other voxels at the edges.

If we choose Face, Edge or Vertex, also voxels connected to at least 10 other voxels over vertex / corner stay. The group of blue voxels shown above is connected to the other voxels at the vertex or corner.