K-Means

Class: NodeKMeansFilter

Node Icon

Classify the intensity values using the K-Means algorithm. This method assigns each voxel a cluster (i.e. label) such that the within-label variance is minimized. This node is useful when the number of clusters are known beforehand, such as classifying CT intensities to Air, Tissue and Bone.

Example Workflows

K-Means clustering

Inputs

Image

Image to be segmented.

Type: Image4DFloat, Required, Single

Outputs

Output

Label map of the same size as the input image.

Type: Image4DFloat

Settings

Use Non Contiguous Labels Boolean

When set to FALSE, the labels are numbered contiguously: {0,1,2..N}. When set to TRUE, the labels are selected in order to span the dynamic range of the output image.

Classes Text

Initial mean of each class, specified as a comma separated list of numbers.

References

ScalarImageKmeansImageFilter in SimpleITK
K-means clustering on Wikipedia