interfaces.semtools.filtering.featuredetection¶
CannyEdge¶
Wraps command ** CannyEdge **
title: Canny Edge Detection
category: Filtering.FeatureDetection
description: Get the distance from a voxel to the nearest voxel of a given tissue type.
version: 0.1.0.(alpha)
documentation-url: http:://www.na-mic.org/
license: https://www.nitrc.org/svn/brains/BuildScripts/trunk/License.txt
contributor: This tool was written by Hans J. Johnson.
Inputs:
[Mandatory]
[Optional]
args: (a unicode string)
Additional parameters to the command
flag: %s
environ: (a dictionary with keys which are a newbytes or None or a
newstr or None and with values which are a newbytes or None or a
newstr or None, nipype default value: {})
Environment variables
ignore_exception: (a boolean, nipype default value: False)
Print an error message instead of throwing an exception in case the
interface fails to run
inputVolume: (an existing file name)
Required: input tissue label image
flag: --inputVolume %s
lowerThreshold: (a float)
Threshold is the lowest allowed value in the output image. Its data
type is the same as the data type of the output image. Any values
below the Threshold level will be replaced with the OutsideValue
parameter value, whose default is zero.
flag: --lowerThreshold %f
outputVolume: (a boolean or a file name)
Required: output image
flag: --outputVolume %s
terminal_output: (u'stream' or u'allatonce' or u'file' or u'none')
Control terminal output: `stream` - displays to terminal immediately
(default), `allatonce` - waits till command is finished to display
output, `file` - writes output to file, `none` - output is ignored
upperThreshold: (a float)
Threshold is the lowest allowed value in the output image. Its data
type is the same as the data type of the output image. Any values
below the Threshold level will be replaced with the OutsideValue
parameter value, whose default is zero.
flag: --upperThreshold %f
variance: (a float)
Variance and Maximum error are used in the Gaussian smoothing of the
input image. See itkDiscreteGaussianImageFilter for information on
these parameters.
flag: --variance %f
Outputs:
outputVolume: (an existing file name)
Required: output image
CannySegmentationLevelSetImageFilter¶
Wraps command ** CannySegmentationLevelSetImageFilter **
title: Canny Level Set Image Filter
category: Filtering.FeatureDetection
description: The CannySegmentationLevelSet is commonly used to refine a manually generated manual mask.
version: 0.3.0
license: CC
contributor: Regina Kim
acknowledgements: This command module was derived from Insight/Examples/Segmentation/CannySegmentationLevelSetImageFilter.cxx (copyright) Insight Software Consortium. See http://wiki.na-mic.org/Wiki/index.php/Slicer3:Execution_Model_Documentation for more detailed descriptions.
Inputs:
[Mandatory]
[Optional]
advectionWeight: (a float)
Controls the smoothness of the resulting mask, small number are more
smooth, large numbers allow more sharp corners.
flag: --advectionWeight %f
args: (a unicode string)
Additional parameters to the command
flag: %s
cannyThreshold: (a float)
Canny Threshold Value
flag: --cannyThreshold %f
cannyVariance: (a float)
Canny variance
flag: --cannyVariance %f
environ: (a dictionary with keys which are a newbytes or None or a
newstr or None and with values which are a newbytes or None or a
newstr or None, nipype default value: {})
Environment variables
ignore_exception: (a boolean, nipype default value: False)
Print an error message instead of throwing an exception in case the
interface fails to run
initialModel: (an existing file name)
flag: --initialModel %s
initialModelIsovalue: (a float)
The identification of the input model iso-surface. (for a binary
image with 0s and 1s use 0.5) (for a binary image with 0s and 255's
use 127.5).
flag: --initialModelIsovalue %f
inputVolume: (an existing file name)
flag: --inputVolume %s
maxIterations: (an integer (int or long))
The
flag: --maxIterations %d
outputSpeedVolume: (a boolean or a file name)
flag: --outputSpeedVolume %s
outputVolume: (a boolean or a file name)
flag: --outputVolume %s
terminal_output: (u'stream' or u'allatonce' or u'file' or u'none')
Control terminal output: `stream` - displays to terminal immediately
(default), `allatonce` - waits till command is finished to display
output, `file` - writes output to file, `none` - output is ignored
Outputs:
outputSpeedVolume: (an existing file name)
outputVolume: (an existing file name)
DilateImage¶
Wraps command ** DilateImage **
title: Dilate Image
category: Filtering.FeatureDetection
description: Uses mathematical morphology to dilate the input images.
version: 0.1.0.$Revision: 1 $(alpha)
documentation-url: http:://www.na-mic.org/
license: https://www.nitrc.org/svn/brains/BuildScripts/trunk/License.txt
contributor: This tool was developed by Mark Scully and Jeremy Bockholt.
Inputs:
[Mandatory]
[Optional]
args: (a unicode string)
Additional parameters to the command
flag: %s
environ: (a dictionary with keys which are a newbytes or None or a
newstr or None and with values which are a newbytes or None or a
newstr or None, nipype default value: {})
Environment variables
ignore_exception: (a boolean, nipype default value: False)
Print an error message instead of throwing an exception in case the
interface fails to run
inputMaskVolume: (an existing file name)
Required: input brain mask image
flag: --inputMaskVolume %s
inputRadius: (an integer (int or long))
Required: input neighborhood radius
flag: --inputRadius %d
inputVolume: (an existing file name)
Required: input image
flag: --inputVolume %s
outputVolume: (a boolean or a file name)
Required: output image
flag: --outputVolume %s
terminal_output: (u'stream' or u'allatonce' or u'file' or u'none')
Control terminal output: `stream` - displays to terminal immediately
(default), `allatonce` - waits till command is finished to display
output, `file` - writes output to file, `none` - output is ignored
Outputs:
outputVolume: (an existing file name)
Required: output image
DilateMask¶
Wraps command ** DilateMask **
title: Dilate Image
category: Filtering.FeatureDetection
description: Uses mathematical morphology to dilate the input images.
version: 0.1.0.$Revision: 1 $(alpha)
documentation-url: http:://www.na-mic.org/
license: https://www.nitrc.org/svn/brains/BuildScripts/trunk/License.txt
contributor: This tool was developed by Mark Scully and Jeremy Bockholt.
Inputs:
[Mandatory]
[Optional]
args: (a unicode string)
Additional parameters to the command
flag: %s
environ: (a dictionary with keys which are a newbytes or None or a
newstr or None and with values which are a newbytes or None or a
newstr or None, nipype default value: {})
Environment variables
ignore_exception: (a boolean, nipype default value: False)
Print an error message instead of throwing an exception in case the
interface fails to run
inputBinaryVolume: (an existing file name)
Required: input brain mask image
flag: --inputBinaryVolume %s
inputVolume: (an existing file name)
Required: input image
flag: --inputVolume %s
lowerThreshold: (a float)
Required: lowerThreshold value
flag: --lowerThreshold %f
outputVolume: (a boolean or a file name)
Required: output image
flag: --outputVolume %s
sizeStructuralElement: (an integer (int or long))
size of structural element. sizeStructuralElement=1 means that 3x3x3
structuring element for 3D
flag: --sizeStructuralElement %d
terminal_output: (u'stream' or u'allatonce' or u'file' or u'none')
Control terminal output: `stream` - displays to terminal immediately
(default), `allatonce` - waits till command is finished to display
output, `file` - writes output to file, `none` - output is ignored
Outputs:
outputVolume: (an existing file name)
Required: output image
DistanceMaps¶
Wraps command ** DistanceMaps **
title: Mauerer Distance
category: Filtering.FeatureDetection
description: Get the distance from a voxel to the nearest voxel of a given tissue type.
version: 0.1.0.$Revision: 1 $(alpha)
documentation-url: http:://www.na-mic.org/
license: https://www.nitrc.org/svn/brains/BuildScripts/trunk/License.txt
contributor: This tool was developed by Mark Scully and Jeremy Bockholt.
Inputs:
[Mandatory]
[Optional]
args: (a unicode string)
Additional parameters to the command
flag: %s
environ: (a dictionary with keys which are a newbytes or None or a
newstr or None and with values which are a newbytes or None or a
newstr or None, nipype default value: {})
Environment variables
ignore_exception: (a boolean, nipype default value: False)
Print an error message instead of throwing an exception in case the
interface fails to run
inputLabelVolume: (an existing file name)
Required: input tissue label image
flag: --inputLabelVolume %s
inputMaskVolume: (an existing file name)
Required: input brain mask image
flag: --inputMaskVolume %s
inputTissueLabel: (an integer (int or long))
Required: input integer value of tissue type used to calculate
distance
flag: --inputTissueLabel %d
outputVolume: (a boolean or a file name)
Required: output image
flag: --outputVolume %s
terminal_output: (u'stream' or u'allatonce' or u'file' or u'none')
Control terminal output: `stream` - displays to terminal immediately
(default), `allatonce` - waits till command is finished to display
output, `file` - writes output to file, `none` - output is ignored
Outputs:
outputVolume: (an existing file name)
Required: output image
DumpBinaryTrainingVectors¶
Wraps command ** DumpBinaryTrainingVectors **
title: Erode Image
category: Filtering.FeatureDetection
description: Uses mathematical morphology to erode the input images.
version: 0.1.0.$Revision: 1 $(alpha)
documentation-url: http:://www.na-mic.org/
license: https://www.nitrc.org/svn/brains/BuildScripts/trunk/License.txt
contributor: This tool was developed by Mark Scully and Jeremy Bockholt.
Inputs:
[Mandatory]
[Optional]
args: (a unicode string)
Additional parameters to the command
flag: %s
environ: (a dictionary with keys which are a newbytes or None or a
newstr or None and with values which are a newbytes or None or a
newstr or None, nipype default value: {})
Environment variables
ignore_exception: (a boolean, nipype default value: False)
Print an error message instead of throwing an exception in case the
interface fails to run
inputHeaderFilename: (an existing file name)
Required: input header file name
flag: --inputHeaderFilename %s
inputVectorFilename: (an existing file name)
Required: input vector filename
flag: --inputVectorFilename %s
terminal_output: (u'stream' or u'allatonce' or u'file' or u'none')
Control terminal output: `stream` - displays to terminal immediately
(default), `allatonce` - waits till command is finished to display
output, `file` - writes output to file, `none` - output is ignored
Outputs:
None
ErodeImage¶
Wraps command ** ErodeImage **
title: Erode Image
category: Filtering.FeatureDetection
description: Uses mathematical morphology to erode the input images.
version: 0.1.0.$Revision: 1 $(alpha)
documentation-url: http:://www.na-mic.org/
license: https://www.nitrc.org/svn/brains/BuildScripts/trunk/License.txt
contributor: This tool was developed by Mark Scully and Jeremy Bockholt.
Inputs:
[Mandatory]
[Optional]
args: (a unicode string)
Additional parameters to the command
flag: %s
environ: (a dictionary with keys which are a newbytes or None or a
newstr or None and with values which are a newbytes or None or a
newstr or None, nipype default value: {})
Environment variables
ignore_exception: (a boolean, nipype default value: False)
Print an error message instead of throwing an exception in case the
interface fails to run
inputMaskVolume: (an existing file name)
Required: input brain mask image
flag: --inputMaskVolume %s
inputRadius: (an integer (int or long))
Required: input neighborhood radius
flag: --inputRadius %d
inputVolume: (an existing file name)
Required: input image
flag: --inputVolume %s
outputVolume: (a boolean or a file name)
Required: output image
flag: --outputVolume %s
terminal_output: (u'stream' or u'allatonce' or u'file' or u'none')
Control terminal output: `stream` - displays to terminal immediately
(default), `allatonce` - waits till command is finished to display
output, `file` - writes output to file, `none` - output is ignored
Outputs:
outputVolume: (an existing file name)
Required: output image
FlippedDifference¶
Wraps command ** FlippedDifference **
title: Flip Image
category: Filtering.FeatureDetection
description: Difference between an image and the axially flipped version of that image.
version: 0.1.0.$Revision: 1 $(alpha)
documentation-url: http:://www.na-mic.org/
license: https://www.nitrc.org/svn/brains/BuildScripts/trunk/License.txt
contributor: This tool was developed by Mark Scully and Jeremy Bockholt.
Inputs:
[Mandatory]
[Optional]
args: (a unicode string)
Additional parameters to the command
flag: %s
environ: (a dictionary with keys which are a newbytes or None or a
newstr or None and with values which are a newbytes or None or a
newstr or None, nipype default value: {})
Environment variables
ignore_exception: (a boolean, nipype default value: False)
Print an error message instead of throwing an exception in case the
interface fails to run
inputMaskVolume: (an existing file name)
Required: input brain mask image
flag: --inputMaskVolume %s
inputVolume: (an existing file name)
Required: input image
flag: --inputVolume %s
outputVolume: (a boolean or a file name)
Required: output image
flag: --outputVolume %s
terminal_output: (u'stream' or u'allatonce' or u'file' or u'none')
Control terminal output: `stream` - displays to terminal immediately
(default), `allatonce` - waits till command is finished to display
output, `file` - writes output to file, `none` - output is ignored
Outputs:
outputVolume: (an existing file name)
Required: output image
GenerateBrainClippedImage¶
Wraps command ** GenerateBrainClippedImage **
title: GenerateBrainClippedImage
category: Filtering.FeatureDetection
description: Automatic FeatureImages using neural networks
version: 1.0
license: https://www.nitrc.org/svn/brains/BuildScripts/trunk/License.txt
contributor: Eun Young Kim
Inputs:
[Mandatory]
[Optional]
args: (a unicode string)
Additional parameters to the command
flag: %s
environ: (a dictionary with keys which are a newbytes or None or a
newstr or None and with values which are a newbytes or None or a
newstr or None, nipype default value: {})
Environment variables
ignore_exception: (a boolean, nipype default value: False)
Print an error message instead of throwing an exception in case the
interface fails to run
inputImg: (an existing file name)
input volume 1, usally t1 image
flag: --inputImg %s
inputMsk: (an existing file name)
input volume 2, usally t2 image
flag: --inputMsk %s
numberOfThreads: (an integer (int or long))
Explicitly specify the maximum number of threads to use.
flag: --numberOfThreads %d
outputFileName: (a boolean or a file name)
(required) output file name
flag: --outputFileName %s
terminal_output: (u'stream' or u'allatonce' or u'file' or u'none')
Control terminal output: `stream` - displays to terminal immediately
(default), `allatonce` - waits till command is finished to display
output, `file` - writes output to file, `none` - output is ignored
Outputs:
outputFileName: (an existing file name)
(required) output file name
GenerateSummedGradientImage¶
Wraps command ** GenerateSummedGradientImage **
title: GenerateSummedGradient
category: Filtering.FeatureDetection
description: Automatic FeatureImages using neural networks
version: 1.0
license: https://www.nitrc.org/svn/brains/BuildScripts/trunk/License.txt
contributor: Greg Harris, Eun Young Kim
Inputs:
[Mandatory]
[Optional]
MaximumGradient: (a boolean)
If set this flag, it will compute maximum gradient between two input
volumes instead of sum of it.
flag: --MaximumGradient
args: (a unicode string)
Additional parameters to the command
flag: %s
environ: (a dictionary with keys which are a newbytes or None or a
newstr or None and with values which are a newbytes or None or a
newstr or None, nipype default value: {})
Environment variables
ignore_exception: (a boolean, nipype default value: False)
Print an error message instead of throwing an exception in case the
interface fails to run
inputVolume1: (an existing file name)
input volume 1, usally t1 image
flag: --inputVolume1 %s
inputVolume2: (an existing file name)
input volume 2, usally t2 image
flag: --inputVolume2 %s
numberOfThreads: (an integer (int or long))
Explicitly specify the maximum number of threads to use.
flag: --numberOfThreads %d
outputFileName: (a boolean or a file name)
(required) output file name
flag: --outputFileName %s
terminal_output: (u'stream' or u'allatonce' or u'file' or u'none')
Control terminal output: `stream` - displays to terminal immediately
(default), `allatonce` - waits till command is finished to display
output, `file` - writes output to file, `none` - output is ignored
Outputs:
outputFileName: (an existing file name)
(required) output file name
GenerateTestImage¶
Wraps command ** GenerateTestImage **
title: DownSampleImage
category: Filtering.FeatureDetection
description: Down sample image for testing
version: 1.0
license: https://www.nitrc.org/svn/brains/BuildScripts/trunk/License.txt
contributor: Eun Young Kim
Inputs:
[Mandatory]
[Optional]
args: (a unicode string)
Additional parameters to the command
flag: %s
environ: (a dictionary with keys which are a newbytes or None or a
newstr or None and with values which are a newbytes or None or a
newstr or None, nipype default value: {})
Environment variables
ignore_exception: (a boolean, nipype default value: False)
Print an error message instead of throwing an exception in case the
interface fails to run
inputVolume: (an existing file name)
input volume 1, usally t1 image
flag: --inputVolume %s
lowerBoundOfOutputVolume: (a float)
flag: --lowerBoundOfOutputVolume %f
outputVolume: (a boolean or a file name)
(required) output file name
flag: --outputVolume %s
outputVolumeSize: (a float)
output Volume Size
flag: --outputVolumeSize %f
terminal_output: (u'stream' or u'allatonce' or u'file' or u'none')
Control terminal output: `stream` - displays to terminal immediately
(default), `allatonce` - waits till command is finished to display
output, `file` - writes output to file, `none` - output is ignored
upperBoundOfOutputVolume: (a float)
flag: --upperBoundOfOutputVolume %f
Outputs:
outputVolume: (an existing file name)
(required) output file name
GradientAnisotropicDiffusionImageFilter¶
Wraps command ** GradientAnisotropicDiffusionImageFilter **
title: GradientAnisopropicDiffusionFilter
category: Filtering.FeatureDetection
description: Image Smoothing using Gradient Anisotropic Diffuesion Filer
contributor: This tool was developed by Eun Young Kim by modifying ITK Example
Inputs:
[Mandatory]
[Optional]
args: (a unicode string)
Additional parameters to the command
flag: %s
conductance: (a float)
Conductance for diffusion process
flag: --conductance %f
environ: (a dictionary with keys which are a newbytes or None or a
newstr or None and with values which are a newbytes or None or a
newstr or None, nipype default value: {})
Environment variables
ignore_exception: (a boolean, nipype default value: False)
Print an error message instead of throwing an exception in case the
interface fails to run
inputVolume: (an existing file name)
Required: input image
flag: --inputVolume %s
numberOfIterations: (an integer (int or long))
Optional value for number of Iterations
flag: --numberOfIterations %d
outputVolume: (a boolean or a file name)
Required: output image
flag: --outputVolume %s
terminal_output: (u'stream' or u'allatonce' or u'file' or u'none')
Control terminal output: `stream` - displays to terminal immediately
(default), `allatonce` - waits till command is finished to display
output, `file` - writes output to file, `none` - output is ignored
timeStep: (a float)
Time step for diffusion process
flag: --timeStep %f
Outputs:
outputVolume: (an existing file name)
Required: output image
HammerAttributeCreator¶
Wraps command ** HammerAttributeCreator **
title: HAMMER Feature Vectors
category: Filtering.FeatureDetection
description: Create the feature vectors used by HAMMER.
version: 0.1.0.$Revision: 1 $(alpha)
documentation-url: http:://www.na-mic.org/
license: https://www.nitrc.org/svn/brains/BuildScripts/trunk/License.txt
contributor: This was extracted from the Hammer Registration source code, and wrapped up by Hans J. Johnson.
Inputs:
[Mandatory]
[Optional]
Scale: (an integer (int or long))
Determine Scale of Ball
flag: --Scale %d
Strength: (a float)
Determine Strength of Edges
flag: --Strength %f
args: (a unicode string)
Additional parameters to the command
flag: %s
environ: (a dictionary with keys which are a newbytes or None or a
newstr or None and with values which are a newbytes or None or a
newstr or None, nipype default value: {})
Environment variables
ignore_exception: (a boolean, nipype default value: False)
Print an error message instead of throwing an exception in case the
interface fails to run
inputCSFVolume: (an existing file name)
Required: input CSF posterior image
flag: --inputCSFVolume %s
inputGMVolume: (an existing file name)
Required: input grey matter posterior image
flag: --inputGMVolume %s
inputWMVolume: (an existing file name)
Required: input white matter posterior image
flag: --inputWMVolume %s
outputVolumeBase: (a unicode string)
Required: output image base name to be appended for each feature
vector.
flag: --outputVolumeBase %s
terminal_output: (u'stream' or u'allatonce' or u'file' or u'none')
Control terminal output: `stream` - displays to terminal immediately
(default), `allatonce` - waits till command is finished to display
output, `file` - writes output to file, `none` - output is ignored
Outputs:
None
NeighborhoodMean¶
Wraps command ** NeighborhoodMean **
title: Neighborhood Mean
category: Filtering.FeatureDetection
description: Calculates the mean, for the given neighborhood size, at each voxel of the T1, T2, and FLAIR.
version: 0.1.0.$Revision: 1 $(alpha)
documentation-url: http:://www.na-mic.org/
license: https://www.nitrc.org/svn/brains/BuildScripts/trunk/License.txt
contributor: This tool was developed by Mark Scully and Jeremy Bockholt.
Inputs:
[Mandatory]
[Optional]
args: (a unicode string)
Additional parameters to the command
flag: %s
environ: (a dictionary with keys which are a newbytes or None or a
newstr or None and with values which are a newbytes or None or a
newstr or None, nipype default value: {})
Environment variables
ignore_exception: (a boolean, nipype default value: False)
Print an error message instead of throwing an exception in case the
interface fails to run
inputMaskVolume: (an existing file name)
Required: input brain mask image
flag: --inputMaskVolume %s
inputRadius: (an integer (int or long))
Required: input neighborhood radius
flag: --inputRadius %d
inputVolume: (an existing file name)
Required: input image
flag: --inputVolume %s
outputVolume: (a boolean or a file name)
Required: output image
flag: --outputVolume %s
terminal_output: (u'stream' or u'allatonce' or u'file' or u'none')
Control terminal output: `stream` - displays to terminal immediately
(default), `allatonce` - waits till command is finished to display
output, `file` - writes output to file, `none` - output is ignored
Outputs:
outputVolume: (an existing file name)
Required: output image
NeighborhoodMedian¶
Wraps command ** NeighborhoodMedian **
title: Neighborhood Median
category: Filtering.FeatureDetection
description: Calculates the median, for the given neighborhood size, at each voxel of the input image.
version: 0.1.0.$Revision: 1 $(alpha)
documentation-url: http:://www.na-mic.org/
license: https://www.nitrc.org/svn/brains/BuildScripts/trunk/License.txt
contributor: This tool was developed by Mark Scully and Jeremy Bockholt.
Inputs:
[Mandatory]
[Optional]
args: (a unicode string)
Additional parameters to the command
flag: %s
environ: (a dictionary with keys which are a newbytes or None or a
newstr or None and with values which are a newbytes or None or a
newstr or None, nipype default value: {})
Environment variables
ignore_exception: (a boolean, nipype default value: False)
Print an error message instead of throwing an exception in case the
interface fails to run
inputMaskVolume: (an existing file name)
Required: input brain mask image
flag: --inputMaskVolume %s
inputRadius: (an integer (int or long))
Required: input neighborhood radius
flag: --inputRadius %d
inputVolume: (an existing file name)
Required: input image
flag: --inputVolume %s
outputVolume: (a boolean or a file name)
Required: output image
flag: --outputVolume %s
terminal_output: (u'stream' or u'allatonce' or u'file' or u'none')
Control terminal output: `stream` - displays to terminal immediately
(default), `allatonce` - waits till command is finished to display
output, `file` - writes output to file, `none` - output is ignored
Outputs:
outputVolume: (an existing file name)
Required: output image
STAPLEAnalysis¶
Wraps command ** STAPLEAnalysis **
title: Dilate Image
category: Filtering.FeatureDetection
description: Uses mathematical morphology to dilate the input images.
version: 0.1.0.$Revision: 1 $(alpha)
documentation-url: http:://www.na-mic.org/
license: https://www.nitrc.org/svn/brains/BuildScripts/trunk/License.txt
contributor: This tool was developed by Mark Scully and Jeremy Bockholt.
Inputs:
[Mandatory]
[Optional]
args: (a unicode string)
Additional parameters to the command
flag: %s
environ: (a dictionary with keys which are a newbytes or None or a
newstr or None and with values which are a newbytes or None or a
newstr or None, nipype default value: {})
Environment variables
ignore_exception: (a boolean, nipype default value: False)
Print an error message instead of throwing an exception in case the
interface fails to run
inputDimension: (an integer (int or long))
Required: input image Dimension 2 or 3
flag: --inputDimension %d
inputLabelVolume: (a list of items which are an existing file name)
Required: input label volume
flag: --inputLabelVolume %s...
outputVolume: (a boolean or a file name)
Required: output image
flag: --outputVolume %s
terminal_output: (u'stream' or u'allatonce' or u'file' or u'none')
Control terminal output: `stream` - displays to terminal immediately
(default), `allatonce` - waits till command is finished to display
output, `file` - writes output to file, `none` - output is ignored
Outputs:
outputVolume: (an existing file name)
Required: output image
TextureFromNoiseImageFilter¶
Wraps command ** TextureFromNoiseImageFilter **
title: TextureFromNoiseImageFilter
category: Filtering.FeatureDetection
description: Calculate the local noise in an image.
version: 0.1.0.$Revision: 1 $(alpha)
documentation-url: http:://www.na-mic.org/
license: https://www.nitrc.org/svn/brains/BuildScripts/trunk/License.txt
contributor: This tool was developed by Eunyoung Regina Kim
Inputs:
[Mandatory]
[Optional]
args: (a unicode string)
Additional parameters to the command
flag: %s
environ: (a dictionary with keys which are a newbytes or None or a
newstr or None and with values which are a newbytes or None or a
newstr or None, nipype default value: {})
Environment variables
ignore_exception: (a boolean, nipype default value: False)
Print an error message instead of throwing an exception in case the
interface fails to run
inputRadius: (an integer (int or long))
Required: input neighborhood radius
flag: --inputRadius %d
inputVolume: (an existing file name)
Required: input image
flag: --inputVolume %s
outputVolume: (a boolean or a file name)
Required: output image
flag: --outputVolume %s
terminal_output: (u'stream' or u'allatonce' or u'file' or u'none')
Control terminal output: `stream` - displays to terminal immediately
(default), `allatonce` - waits till command is finished to display
output, `file` - writes output to file, `none` - output is ignored
Outputs:
outputVolume: (an existing file name)
Required: output image
TextureMeasureFilter¶
Wraps command ** TextureMeasureFilter **
title: Canny Level Set Image Filter
category: Filtering.FeatureDetection
description: The CannySegmentationLevelSet is commonly used to refine a manually generated manual mask.
version: 0.3.0
license: CC
contributor: Regina Kim
acknowledgements: This command module was derived from Insight/Examples/Segmentation/CannySegmentationLevelSetImageFilter.cxx (copyright) Insight Software Consortium. See http://wiki.na-mic.org/Wiki/index.php/Slicer3:Execution_Model_Documentation for more detailed descriptions.
Inputs:
[Mandatory]
[Optional]
args: (a unicode string)
Additional parameters to the command
flag: %s
distance: (an integer (int or long))
flag: --distance %d
environ: (a dictionary with keys which are a newbytes or None or a
newstr or None and with values which are a newbytes or None or a
newstr or None, nipype default value: {})
Environment variables
ignore_exception: (a boolean, nipype default value: False)
Print an error message instead of throwing an exception in case the
interface fails to run
inputMaskVolume: (an existing file name)
flag: --inputMaskVolume %s
inputVolume: (an existing file name)
flag: --inputVolume %s
insideROIValue: (a float)
flag: --insideROIValue %f
outputFilename: (a boolean or a file name)
flag: --outputFilename %s
terminal_output: (u'stream' or u'allatonce' or u'file' or u'none')
Control terminal output: `stream` - displays to terminal immediately
(default), `allatonce` - waits till command is finished to display
output, `file` - writes output to file, `none` - output is ignored
Outputs:
outputFilename: (an existing file name)