Laws' textures - cerr/CERR GitHub Wiki

Laws textures are computed by convolutions with small (3x3 or 5x5) integer coefficient masks. Laws Energy maps are computed using a moving-window absolute average operation on Laws texture maps [ 1 ].

Available filters

Law's convolution kernels are generated from the following set of 1-D convolution kernels of lengths 3 and 5, as defined in [ 2 ].

1. Level (average gray level)
L3 = [ 1, 2, 1]
L5 = [ 1, 4, 6, 4, 1]

2. Edge (extract edge features)
E3 = [-1, 0, 1]
E5 = [-1, -2, 0, 2, 1]

3. Spot (extract spots)
S3 = [-1, 2, -1]
S5 = [-1, 0, 2, 0, -1]

4. Wave (extract wave features)
W5 = [-1, 2, 0, -2, 1]

5. Ripple (extract ripples)
R5 = [ 1, -4, 6, -4, 1]

CERR also supports Laws Energy maps, computed as the local mean of absolute intensities in a Laws texture map.

Implementation

2-D & 3-D convolution kernels are computed by combining these 1-D convolution kernels using outer products.
CERR provides the 2-D and 3-D kernels generated using all possible combinations of these 1-D masks. Kernels are normalized if the normFlag input is set to 'Yes' (required for IBSI compliance).

Filter parameters

Laws' convolution
Direction.: 1- 2D 2- 3D 3- All
Type : 1- 3, 2- 5, 3- All
Normalize: 'Yes', 'No'
Laws energy
Direction: 1- 2D 2- 3D 3- All
Type: 1- 3, 2- 5, 3- All
Normalize: 'Yes', 'No'
EnergyKernelSize: 3x1 vector specifying mean filter dimensions (set last element to 0 for 2D)
EnergyPadMethod: Padding method for mean filter
EnergyPadSize: 3x1 vector specifying padding dimensions for mean filter (set last element to 0 for 2D)

Usage

Filters can be applied by manually creating a dictionary of parameters or passing them via JSON-format configuration files.

Dictionaries are created using the syntax:

paramS.<parameter_name>.val = <value>;

where parameter_name is one of the supported filter parameters listed above and value, one of the associated arguments.

Sample filter configuration

The following settings can be used to (1) apply the L5E5 Laws Energy filter with normalization after cropping around the specified structure and padding by 2 voxel widths and (2) calculate first-order statistics from the filtered image.

{

"structures": ["ROI"],

"imageType": {
  		 "LawsEnergy": {
                "Direction": "2d",
                "Type": "L5E5",
		"Normalize": "yes",
                "EnergyKernelSize": [15,15],
                "EnergyPadMethod": "mirror",
                "EnergyPadSize" : [7,7,0]
	    },

"settings": {		   
                
                   "padding": {
                                "cropToMaskBounds": "yes"
                                "method": "mirror",  
                                "size": [2,2,0],
			       },

                   "firstOrder" :{ 
				     "offsetForEnergy": 0,
                                     "binNumEntropy":0
                                 }

            }
}

Configurations are parsed and converted to parameter dictionaries using

paramFile = 'Path/to/config_file.json'
paramS = getRadiomicsParamTemplate(paramFile);

Generating filtered images

global planC
strNum = 1;
scanNum = [];  % leave empty when passing structNum to automatically identify associated scan index
planC = generateTextureMapFromPlanC(planC,scanNum,strNum,paramS)

Extracting radiomic features

global planC
scanNum  = 1;
strNum = 2; 
featureS = calcGlobalRadiomicsFeatures(scanNum, structNum, paramS, planC);

References

[1] Rodojevic, G. L. M. Texture segmentation: Co-occurrence matrix and Laws’ texture masks methods. Energy, 1, 1. [2] IBSI-2 [https://www.overleaf.com/project/5da9e0b82f399f0001ad3970]