numpyradiomics.glcm#

numpyradiomics.glcm(image, mask, binWidth=25, distances=[1], symmetricalGLCM=True, weightingNorm=None)[source]#

Compute 24 Pyradiomics-style GLCM (Gray Level Co-occurrence Matrix) features.

The GLCM describes the second-order joint probability function of an image region, constrained by the mask. It counts how often pairs of voxels with specific gray levels and specific spatial relationships occur.

Parameters:

  • image (np.ndarray) – 3D image array containing voxel intensities.

  • mask (np.ndarray) – 3D mask array (same shape as image), where non-zero values indicate the ROI.

  • binWidth (float, optional) – Width of bins for discretization. Default is 25.

  • distances (list, optional) – List of integer distances (offsets) to compute GLCMs for. Default is [1].

  • symmetricalGLCM (bool, optional) – If True, counts co-occurrences in both directions (i->j and j->i). If False, only counts i->j. Default is True.

  • weightingNorm (str, optional) – Method to weight GLCMs if multiple distances are used. Options: None, ‘manhattan’, ‘euclidean’, ‘infinity’. Default is None.

Returns:

Dictionary containing the 24 GLCM features:

  • Autocorrelation: Magnitude of the fineness and coarseness of texture.

  • ClusterProminence: Asymmetry of the GLCM.

  • ClusterShade: Skewness of the matrix.

  • ClusterTendency: Grouping of voxels with similar gray-level values.

  • Contrast: Local intensity variation (favors contributions away from the diagonal).

  • Correlation: Linear dependency of gray levels to their neighbors.

  • DifferenceAverage: Mean of the difference distribution.

  • DifferenceEntropy: Randomness/variability in neighborhood intensity differences.

  • DifferenceVariance: Variance of the difference distribution.

  • JointAverage: Mean gray level intensity of the $i$ distribution.

  • JointEnergy: Homogeneity of the texture (sum of squared elements).

  • JointEntropy: Randomness/variability in neighborhood intensity values.

  • Id: Inverse Difference (homogeneity 1).

  • Idm: Inverse Difference Moment (homogeneity 2).

  • Idmn: Normalized Inverse Difference Moment.

  • Idn: Normalized Inverse Difference.

  • Imc1: Informational Measure of Correlation 1.

  • Imc2: Informational Measure of Correlation 2.

  • InverseVariance: Inverse variance of the difference distribution.

  • MCC: Maximal Correlation Coefficient.

  • MaximumProbability: Occurrence of the most predominant pair of neighboring intensity values.

  • SumAverage: Mean of the sum distribution.

  • SumEntropy: Sum of entropies.

  • SumSquares: Variance of the distribution (Cluster Tendency is a specific case of this).

Return type:

dict

Example

>>> import numpyradiomics as npr
>>> # Generate a noisy ellipsoid
>>> img, mask = npr.dro.noisy_ellipsoid(radii_mm=(10, 10, 10), intensity_range=(0, 100))
>>>
>>> # Compute GLCM features (distance=1)
>>> feats = npr.glcm(img, mask, binWidth=10, distances=[1])
>>>
>>> print(f"Contrast: {feats['Contrast']:.4f}")
Contrast: 12.4501
On this page

This Page