Segmentation properties

When working on a local dataset that has segmeted images in the MorphoNet Standalone, segmentation segmentation properties are automatically computed.

Using properties during the curation of a segmented dataset are a central part of the quality assessment of a segmentation.

You will see them in the Properties tab, and they are recognizable by the scikit-image logo:

Two of these properties are automatically computed by default : volume and bbox. The other ones are computed on demand, and once thay have been asked by the user, they will continuously be computed at each curation step. Please note that asking for a lot of region properties can slow down the loading of the dataset at each subsequent curation step.

Unless specified, properties are computed on the image scaled by the voxel size, to give accurate physical measurements.

Here you will get a description of these Properties.

Description of the properties

The majority of the region properties are computed from the scikit-image python package, as described here. Properties marked with a (v) are computed on the "original image", whereas others are computed on an image re-scaled by its voxel size, to give appropriate physical measurements.

  • volume: (v) Corresponds to the area property of scikit-image. Area of the region, i.e. number of voxels in the region.

  • volume-real: Volume property but scaled by the voxel-size to give a real physical volume

  • volume-bbox: Corresponds to the area_bbox property of scikit-image. Area of the bounding box i.e. number of voxels of bounding box scaled by voxel-area.

  • volume-filled: Corresponds to the area_filled property of scikit-image. Area of the region with all holes filled in.

  • axis-major-length: The length of the major axis of the ellipse that has the same normalized second central moments as the region.

  • axis-minor-length: The length of the minor axis of the ellipse that has the same normalized second central moments as the region.

  • axis-ratio: Ratio of the longest axis over the smallest axis of the label (see axis-major-length and axis-minor-length).

  • diameter: (v) The mean diameter of the region. It is the mean of axis-major-length and axis-minor-length. Computed as (axis-major-length + axis-minor-length) / 2. IMPORTANT: this value is exprimed in voxels, not physical size, so it can be of use for plugins that use voxel measurements, such as Cellpose for instance.

  • equivalent-diameter-area: The diameter of a circle with the same volume as the region.

  • euler-number: Euler characteristic of the set of non-zero pixels. Computed as number of connected components subtracted by number of holes (input.ndim connectivity). In 3D, number of connected components plus number of holes subtracted by number of tunnels.

  • extent: Ratio of pixels in the region to pixels in the total bounding box. Computed as area / (rows * cols)

  • connected-neighbors: number of connected other labels.

  • convexity: distance to convexity. Computed as volume of the convex hull (smallest convex polygon that encloses the region) divided by the volume of the region.

  • roughness: mean of the absolute values of the region minus the closing (dilation following by erosion) of the region.

  • compactness: computed as (region surface area ** 3) / (36 * pi * (volume ** 2))

  • smoothness: computed as region surface area / volume ** (2 / 3))

  • intensity-max: (v) Value with the greatest intensity in the region.

  • intensity-mean: (v) Value with the mean intensity in the region.

  • intensity-min: (v) Value with the least intensity in the region.

  • intensity-border-variation: (v) the standard deviation of the intensity images only at the border of the segmentation.

  • intensity-offset: (v) the euclidian distance between the gravity center of the intensity images and the geometrical center of the segmentation.

  • lineage-distance : Requires a lineage property, as well as the cell_name property, containing the Conklin naming of cells. Represents the tree-edit distance between lineage trees of symmetrical cells.

  • bbox: (v) bounding box of the region (tuple).