Table 2.

Overview of textural features with definitions subdivided into textural features of 1st and 2nd-order

1st-order textural features
Heterogeneity	= presence of edges detected by the use of a Laplacian of Gaussian filter
Intensity	= texture intensity as the voxel value of the corresponding input image voxel
Average	= noise independent voxel intensity
Deviation	= correlates with the local range of input image voxel values
Skewness	= describes if the current neighbourhood has a centered distribution of grey values
2nd-order textural features
Entropy of co-occurrence matrix	= entropy of the distribution of two co-occurring neighbour grey values
Number non-uniformity (NGLDM)	= the sum of squared NGLDM matrix elements divided by the sum of (unsquared) matrix elements
Entropy of NGLDM	= considers NGLDM matrix entries as random variables with an underlying statistical distribution, an image with a certain kind of regularity
Entropy of heterogeneity	= the randomness on the presence and distribution of edges
Entropie (NGLDM)	= considering NGLDM matrix entries as random variables with an underlying statistical distribution, an image with a certain kind of regularity
Contrast (NGTDM)	= correlation of grey value differences between neighbouring voxels (DifferencegreyValueNeigbors) with the range of voxels in the whole neighbourhood of the current voxel (Rangeneighborhood). The texture value for the current voxel is computed as: textureValuecurrentVoxel = Rangeneighborhood * DifferencegreyValueNeigbors

Abbreviations: NGLDM Neighbouring Grey-Level Dependence Matrix