NGLDM_global_features - cerr/CERR GitHub Wiki

The feature definitions are available in IBSI section 3.11

List of available features

1. Low dependence emphasis (lde) emphasizes low neighboring grey level dependence counts. A higher value indicates smaller dependence and less homogeneous textures.

$\frac{\sum^{N_g}_{i=1}\sum^{N_d}_{j=1}{\frac{\textbf{P}(i,j)}{i^2}}}{N_z}$

2. High dependence emphasis (hde) emphasizes high neighboring grey level dependence counts. A higher value indicates larger dependence and more homogeneous textures.

$\frac{\sum^{N_g}_{i=1}\sum^{N_d}_{j=1}{\textbf{P}(i,j)j^2}}{N_z}$

3. Low grey level count emphasis (lgce) grey level analogue to low dependence emphasis . Instead of low neighboring grey level dependence counts, low grey levels are emphasized. A higher value indicates a greater concentration of low gray-level values in the image.

$\frac{\sum^{N_g}_{i=1}\sum^{N_d}_{j=1}{\frac{\textbf{P}(i,j)}{i^2}}}{N_z}$

4. High grey level count emphasis (hgce) is a grey level analogue to high dependence emphasis. The feature emphasizes high grey levels. A higher value indicates a greater concentration of high gray-level values in the image.

$\frac{\sum^{N_g}_{i=1}\sum^{N_d}_{j=1}{\textbf{P}(i,j)i^2}}{N_z}$

5. Low dependence low grey level emphasis (ldlge) emphasizes neighboring grey level dependence counts in the upper left quadrant of the NGLDM, where low dependence counts and low grey levels are located. Measures the joint distribution of small dependence with lower gray-level values.

$\frac{\sum^{N_g}_{i=1}\sum^{N_d}_{j=1}{\frac{\textbf{P}(i,j)}{i^2j^2}}}{N_z}$

6. Low dependence high grey level emphasis (ldhge) emphasises neighbouring grey level dependence counts in the lower left quadrant of the NGLDM, where low dependence counts and high grey levels are located. Measures the joint distribution of small dependence with higher gray-level values.

$\frac{\sum^{N_g}_{i=1}\sum^{N_d}_{j=1}{\frac{\textbf{P}(i,j)i^2}{j^2}}}{N_z}$

7. High dependence low grey level emphasis (hdlge) emphasizes neighboring grey level dependence counts in the upper right quadrant of the NGLDM, where high dependence counts and low grey levels are located. Measures the joint distribution of large dependence with lower gray-level values.

$\frac{\sum^{N_g}_{i=1}\sum^{N_d}_{j=1}{\frac{\textbf{P}(i,j)j^2}{i^2}}}{N_z}$

8. High dependence high grey level emphasis (hdhge) emphasizes neighboring grey level dependence counts in the lower right quadrant of the NGLDM, where high dependence counts and high grey levels are located. Measures the joint distribution of large dependence with higher gray-level values.

$\frac{\sum^{N_g}_{i=1}\sum^{N_d}_{j=1}{\textbf{P}(i,j)i^2j^2}}{N_z}$

9. Grey level non-uniformity (gln) measures the distribution of neighboring grey level dependence counts over the grey values. The feature value is low when dependence counts are equally distributed along grey levels. A lower value correlates with a greater similarity in intensity values.

$\frac{\sum^{N_g}_{i=1}\left(\sum^{N_d}_{j=1}{\textbf{P}(i,j)}\right)^2}{N_z}$

10. Grey level non-uniformity normalized (glnNorm) normalized version of the grey level non-uniformity feature. It is mathematically equivalent to the First Order Uniformity.

$\frac{\sum^{N_g}_{i=1}\left(\sum^{N_d}_{j=1}{\textbf{P}(i,j)}\right)^2}{\sum^{N_g}_{i=1}\sum^{N_d}_{j=1}{\textbf{P}(i,j)}^2}$

11. Dependence count non-uniformity (dcn) assesses the distribution of neighbouring grey level dependence counts over the different dependence counts. The feature value is low when dependence counts are equally distributed.

$\frac{\sum^{N_d}_{j=1}\left(\sum^{N_g}_{i=1}{\textbf{P}(i,j)}\right)^2}{N_z}$

12. Dependence count non-uniformity normalised (dcnNorm) normalized version of the dependence count non-uniformity feature.

$\frac{\sum^{N_d}_{j=1}\left(\sum^{N_g}_{i=1}{\textbf{P}(i,j)}\right)^2}{N_z^2}$

13. Dependence count percentage (dcp) assesses the fraction of the number of realized neighborhoods and the maximum number of potential neighborhoods. Dependence count percentage may be completely omitted as it evaluates to 1 when complete neighborhoods are not required, as is the case under our definition.

$\frac{N_z}{N_p}$

14. Grey level variance (glv)

$\displaystyle\sum^{N_g}_{i=1}\displaystyle\sum^{N_d}_{j=1}{p(i,j)(i - \mu)^2} \text{, where} \mu = \displaystyle\sum^{N_g}_{i=1}\displaystyle\sum^{N_d}_{j=1}{ip(i,j)}$

15. Dependence count variance (dcv)

$\displaystyle\sum^{N_g}_{i=1}\displaystyle\sum^{N_d}_{j=1}{p(i,j)(j - \mu)^2} \text{, where} \mu = \displaystyle\sum^{N_g}_{i=1}\displaystyle\sum^{N_d}_{j=1}{jp(i,j)}$

16. Dependence count entropy (dcEntropy)

$-\displaystyle\sum^{N_g}_{i=1}\displaystyle\sum^{N_d}_{j=1}{p(i,j)\log_{2}(p(i,j)+\epsilon)}$

17. Dependence count energy (dcEnergy)

$-\displaystyle\sum^{N_g}_{i=1}\sum^{N_d}_{j=1}{p(i,j)^2}$

where
is the total number of gray levels in the image.
is the number of dependency sizes in the image.
is the number of dependency zones in the image.
$p(i,j) = \frac{\textbf{P}(i,j)}{N_z}$ is the normalized dependence matrix.

Calculation details

Let quantizedM be the discretized scan matrix. The number of voxels in the region of interest is calculated as

numVoxels = sum(~isnan(quantizedM(:)));

The NGLDM can be built using 2-d or 3-d neighborhoods and with various intensity difference thresholds. The patchRadius3dV and imgDiffThresh argument to the calcNGLDM function controls this.

patchRadius3dV = [1 1 1]; % [1 1 1]: 3d, [1 1 0]: 2d
imgDiffThresh = 1;

The NGLDM is computed as follows:

ngldmM = calcNGLDM(quantizedM, patchRadius3dV, numGrLevels, imgDiffThresh);

The scalar features from the NGLDM matrix are obtained as follows:

featureS = ngldmToScalarFeatures(ngldmM,numVoxels);

NGLDM_global_features - cerr/CERR GitHub Wiki

List of available features

Calculation details

⚠️ **GitHub.com Fallback** ⚠️

⚠️ GitHub.com Fallback ⚠️