Slicer3:Execution Model Documentation:Python
From NAMIC Wiki
Home < Slicer3:Execution Model Documentation:Python
Contents
Python Modules
A Python interpreter has been integrated into Slicer3. More details of the Python language can be found at Python.org. This interpreter may be used to execute Python script plugins. The modules have a few specific requirements to be correctly found and used as plugins. Like the standard executable and shared library plugins, Python plugins need to be self describing. To do this, the script must have a top level variable called XML or provide a toXML procedure. For example:
XML = """<?xml version="1.0" encoding="utf-8"?> <executable> <category>Filtering.Denoising</category> ... def toXML(): return XML;
Details of the XML format are found in the main Execution Model documentation. Rather than construct a command line to pass into Python, Slicer3 directly calls an Execute procedure. It is assumed that the Execute function expects positional arguments first, and any optional arguments are passed in as keyword arguments. For instance, the GradientAnisotropicDiffusion.py module provides an Execute:
def Execute ( inputVolume, outputVolume, conductance=1.0, timeStep=0.0625, iterations=1 ):
print "Executing Python Demo Application!"
Slicer = __import__ ( "Slicer" );
slicer = Slicer.Slicer()
in = slicer.MRMLScene.GetNodeByID ( inputVolume );
out = slicer.MRMLScene.GetNodeByID ( outputVolume );
filter = slicer.vtkITKGradientAnisotropicDiffusionImageFilter.New()
filter.SetConductanceParameter ( conductance )
filter.SetTimeStep ( timeStep )
filter.SetNumberOfIterations ( iterations )
filter.SetInput ( in.GetImageData() )
filter.Update()
out.SetAndObserveImageData(filter.GetOutput())
return
The function first constructs a Slicer object by importing the Slicer module. The Slicer object is the main interface into Slicer3 as a whole. The first two arguments, inputVolume and outputVolume are not proper MRMLVolumes, and must be looked up using the Slicer object (that is, the arguments are unique IDs rather than pointers to the objects). The filter is constructed through the Slicer object, and the parameters are set. After the filter is updated, the output image is put using the SetAndObserveImageData method on the output volume.
ToDo
- Progress functionality
- Casting image arguments to proper MRML volume objects before calling Execute
Python Module collection
You can download the zipfile with all these files and (some work in progress) here: File:PythonModules.zip. Either place all the files in the Slicer3/Modules/Python directory (replacing the existing files) or just throw the .py files you're interested in into Slicer3-build/lib/Slicer3/Plugins directory and launch Slicer3.
PythonScript.py
This module allows to run external Python code saved in a .py file on an image or surface (or both), and have the results loaded back in Slicer. It doesn't really do anything itself, but it's handy for:
- writing Python modules, since one can store the actual module code in a file without having to copy it to Slicer3-build/lib/Slicer3/Plugins
- writing quick one-shot filtering operations that are not elegible to become full-fledged modules
- experimenting with Python in Slicer.
XML = """<?xml version="1.0" encoding="utf-8"?>
<executable>
<category>Python Modules</category>
<title>Python Script</title>
<description>Run external Python code on a surface or image.</description>
<version>1.0</version>
<documentation-url></documentation-url>
<license></license>
<contributor>Luca Antiga and Daniel Blezek</contributor>
<parameters>
<label>Python Script Parameters</label>
<description>Parameters for Python script</description>
<file>
<name>scriptFileName</name>
<longflag>scriptFileName</longflag>
<description>File containing Python code to run</description>
<label>Script file</label>
</file>
</parameters>
<parameters>
<label>IO</label>
<description>Input/output parameters</description>
<geometry>
<name>inputSurface</name>
<longflag>inputSurface</longflag>
<label>Input Surface</label>
<channel>input</channel>
<description>Input surface to be filtered</description>
</geometry>
<image>
<name>inputVolume</name>
<longflag>inputVolume</longflag>
<label>Input Volume</label>
<channel>input</channel>
<description>Input image to be filtered</description>
</image>
<geometry>
<name>outputSurface</name>
<longflag>outputSurface</longflag>
<label>Output Surface</label>
<channel>output</channel>
<description>Output filtered surface</description>
</geometry>
<image>
<name>outputVolume</name>
<longflag>outputVolume</longflag>
<label>Output Volume</label>
<channel>output</channel>
<description>Output filtered volume</description>
</image>
</parameters>
</executable>
"""
def Execute (scriptFileName="", inputSurface="", inputVolume="", outputSurface="", outputVolume=""):
Slicer = __import__("Slicer")
slicer = Slicer.Slicer()
scene = slicer.MRMLScene
if inputSurface:
inputSurface = scene.GetNodeByID(inputSurface)
if inputVolume:
inputVolume = scene.GetNodeByID(inputVolume)
if outputSurface:
outputSurface = scene.GetNodeByID(outputSurface)
if outputVolume:
outputVolume = scene.GetNodeByID(outputVolume)
try:
execfile(scriptFileName)
except Exception, error:
slicer.Application.ErrorMessage("Python script error: %s" % error)
return
Note how the try...except block intercepts Python exceptions generated while running the external script and prints messages out in Slicer's message console.
Now let's see what an external script code could look like. Say we want to take the input surface and decimate it.
inputPolyData = inputSurface.GetPolyData() decimation = slicer.vtkDecimatePro.New() decimation.SetInput(inputSurface.GetPolyData()) decimation.SetTargetReduction(0.99) decimation.Update() outputSurface.SetAndObservePolyData(decimation.GetOutput())
In the script (which we can save in a .py file), we don't have to start by importing Slicer module, since it's already done in the PythonScript.py module. We have to assume that inputSurface, outputSurface, inputVolume and outputVolume are already available as MRML nodes (vtkMRMLModelNode and vtkMRMLVolumeNode, respectively), and that the slicer module is also already imported.
Setting of the filtered poly data to the outputSurface is left to the user (see last line).
In general, the syntax used to use VTK classes in a Slicer Python module is very similar to using VTK from Python. There are just two main differences:
- classes have to looked for in the slicer object
- classes have to be instantiated using .New(), instead of the Python-type constructor
In practice
polydata = slicer.vtkPolyData.New()
instead of
polydata = vtk.vtkPolyData()
SurfaceConnectivity.py
This module runs a connectivity filter on an input surface. There are two modes of operation: one is to extract all regions and color them accordingly, the other is to extract the region closest to a fiducial. Optionally, the module will generate a list of fiducials named "Region #" placed on a point of each connected region. It's a nice example on how one can interact with the MRML scene. FIXME: since Python modules are run in the processing thread, AddNode currently causes a crash (possible solution - provide a thread-safe version of it).
XML = """<?xml version="1.0" encoding="utf-8"?>
<executable>
<category>Python Modules</category>
<title>Python Surface Connectivity</title>
<description>
Identify connected regions of a surface.
</description>
<version>1.0</version>
<documentation-url></documentation-url>
<license></license>
<contributor>Luca Antiga and Daniel Blezek</contributor>
<parameters>
<label>Surface Connectivity Parameters</label>
<description>Parameters for surface connectivity</description>
<string-enumeration>
<name>connectivityMode</name>
<longflag>connectivityMode</longflag>
<description>Mode of operation of connectivity filter</description>
<label>Connectivity mode</label>
<default>AllRegions</default>
<element>AllRegions</element>
<element>ClosestToSeed</element>
</string-enumeration>
<boolean>
<name>enableOutputFiducials</name>
<longflag>enableOutputFiducials</longflag>
<description>Toggle generation of labeled fiducials corresponding to each extracted region</description>
<label>Enable output fiducials</label>
<default>false</default>
</boolean>
<point multiple="false">
<name>seedPoint</name>
<longflag>seedPoint</longflag>
<label>Seed Point</label>
<description>Fiducial to extract closest surface</description>
</point>
</parameters>
<parameters>
<label>IO</label>
<description>Input/output parameters</description>
<geometry>
<name>inputSurface</name>
<label>Input Surface</label>
<channel>input</channel>
<index>0</index>
<description>Input surface to be filtered</description>
</geometry>
<geometry>
<name>outputSurface</name>
<label>Output Surface</label>
<channel>output</channel>
<index>1</index>
<description>Output filtered surface</description>
</geometry>
</parameters>
</executable>
"""
def Execute (inputSurface, outputSurface, connectivityMode="AllRegions", enableOutputFiducials=False, seedPoint=[0.0,0.0,0.0]):
Slicer = __import__("Slicer")
slicer = Slicer.Slicer()
scene = slicer.MRMLScene
inputSurface = scene.GetNodeByID(inputSurface)
outputSurface = scene.GetNodeByID(outputSurface)
connectivityFilter = slicer.vtkPolyDataConnectivityFilter.New()
connectivityFilter.SetInput(inputSurface.GetPolyData())
if connectivityMode == "AllRegions":
connectivityFilter.SetExtractionModeToAllRegions()
connectivityFilter.ColorRegionsOn()
elif connectivityMode == "ClosestToSeed":
connectivityFilter.SetExtractionModeToClosestPointRegion()
connectivityFilter.SetClosestPoint(seedPoint[0],seedPoint[1],seedPoint[2])
connectivityFilter.Update()
if enableOutputFiducials:
fiducialList = scene.CreateNodeByClass("vtkMRMLFiducialListNode")
fiducialList.DisableModifiedEventOn()
fiducialList.SetScene(scene)
thresholdPoints = slicer.vtkThresholdPoints.New()
thresholdPoints.SetInput(connectivityFilter.GetOutput())
numberOfRegions = connectivityFilter.GetNumberOfExtractedRegions()
for i in range(numberOfRegions):
lower = i - 0.5
upper = i + 0.5
thresholdPoints.ThresholdBetween(lower,upper)
thresholdPoints.Update()
if thresholdPoints.GetOutput().GetNumberOfPoints() > 0:
point = thresholdPoints.GetOutput().GetPoint(0)
fid = fiducialList.AddFiducial()
fiducialList.SetNthFiducialXYZ(fid,point[0],point[1],point[2])
fiducialList.SetNthFiducialLabelText(fid,"Region %d" % i)
slicer.ApplicationLogic.RequestModified(fiducialList)
fiducialList.DisableModifiedEventOff()
#FIXME: this causes a crash
scene.AddNode(fiducialList)
outputSurface.SetAndObservePolyData(connectivityFilter.GetOutput())
return
SurfaceICPRegistration.py
This module provides surface registration using the Iterative Closest Point algorithm as implemented in VTK. Nothing fancy going on, just nice functionality added to Slicer.
XML = """<?xml version="1.0" encoding="utf-8"?>
<executable>
<category>Python Modules</category>
<title>Python Surface ICP Registration</title>
<description>
Performes registration of the input surface onto a target surface using on the Iterative Closest Point algorithm.
</description>
<version>1.0</version>
<documentation-url></documentation-url>
<license></license>
<contributor>Luca Antiga and Daniel Blezek</contributor>
<parameters>
<label>Surface ICP Registration Parameters</label>
<description>Parameters for surface registration</description>
<string-enumeration>
<name>landmarkTransformMode</name>
<longflag>landmarkTransformMode</longflag>
<description>Mode of operation of internal landmark transform</description>
<label>Landmark transform mode</label>
<default>RigidBody</default>
<element>RigidBody</element>
<element>Similarity</element>
<element>Affine</element>
</string-enumeration>
<string-enumeration>
<name>meanDistanceMode</name>
<longflag>meanDistanceMode</longflag>
<label>Mean distance mode</label>
<description>Set how the mean distance should be computed, RMS (square root of the average of the sum of squares of the closest point distances) or AbsoluteValue (mean of the sum of absolute values of the closest point distances)</description>
<default>RMS</default>
<element>RMS</element>
<element>AbsoluteValue</element>
</string-enumeration>
<integer>
<name>maximumNumberOfIterations</name>
<longflag>maximumNumberOfIterations</longflag>
<label>Maximum number of iterations</label>
<description>Maximum number of Interative Closest Point iterations</description>
<default>50</default>
<constraints>
<minimum>1</minimum>
<maximum>1000</maximum>
<step>1</step>
</constraints>
</integer>
<integer>
<name>maximumNumberOfLandmarks</name>
<longflag>maximumNumberOfLandmarks</longflag>
<label>Maximum number of landmarks</label>
<description>Maximum number of landmarks on source and target surfaces</description>
<default>200</default>
<constraints>
<minimum>1</minimum>
<maximum>1000</maximum>
<step>1</step>
</constraints>
</integer>
<boolean>
<name>startByMatchingCentroids</name>
<longflag>startByMatchingCentroids</longflag>
<label>Start by matching centroids</label>
<description>Toggle starting registration by first translating the source centroid to the target centroid</description>
<default>false</default>
</boolean>
<boolean>
<name>checkMeanDistance</name>
<longflag>checkMeanDistance</longflag>
<label>Check mean distance</label>
<description>Force the algorithm to check the mean distance between two iterations</description>
<default>false</default>
</boolean>
<double>
<name>maximumMeanDistance</name>
<longflag>maximumMeanDistance</longflag>
<label>Maximum mean distance</label>
<description>Maximum distance between two iterations used for determining convergence</description>
<default>0.01</default>
</double>
</parameters>
<parameters>
<label>IO</label>
<description>Input/output parameters</description>
<geometry>
<name>inputSurface</name>
<label>Input Surface</label>
<channel>input</channel>
<index>0</index>
<description>Input surface to be registered</description>
</geometry>
<geometry>
<name>targetSurface</name>
<label>Target Surface</label>
<channel>input</channel>
<index>1</index>
<description>Target surface for registration</description>
</geometry>
<geometry>
<name>outputSurface</name>
<label>Output Surface</label>
<channel>output</channel>
<index>2</index>
<description>Output registered surface</description>
</geometry>
</parameters>
</executable>
"""
def Execute (inputSurface, targetSurface, outputSurface, \
landmarkTransformMode="RigidBody", meanDistanceMode="RMS", \
maximumNumberOfIterations=50, maximumNumberOfLandmarks=200, \
startByMatchingCentroids=False, checkMeanDistance=False, maximumMeanDistance=0.01):
Slicer = __import__("Slicer")
slicer = Slicer.Slicer()
scene = slicer.MRMLScene
inputSurface = scene.GetNodeByID(inputSurface)
targetSurface = scene.GetNodeByID(targetSurface)
outputSurface = scene.GetNodeByID(outputSurface)
icpTransform = slicer.vtkIterativeClosestPointTransform.New()
icpTransform.SetSource(inputSurface.GetPolyData())
icpTransform.SetTarget(targetSurface.GetPolyData())
if landmarkTransformMode == "RigidBody":
icpTransform.GetLandmarkTransform().SetModeToRigidBody()
elif landmarkTransformMode == "Similarity":
icpTransform.GetLandmarkTransform().SetModeToSimilarity()
elif landmarkTransformMode == "Affine":
icpTransform.GetLandmarkTransform().SetModeToAffine()
if meanDistanceMode == "RMS":
icpTransform.SetMeanDistanceModeToRMS()
elif meanDistanceMode == "AbsoluteValue":
icpTransform.SetMeanDistanceModeToAbsoluteValue()
icpTransform.SetMaximumNumberOfIterations(maximumNumberOfIterations)
icpTransform.SetMaximumNumberOfLandmarks(maximumNumberOfLandmarks)
icpTransform.SetStartByMatchingCentroids(int(startByMatchingCentroids))
icpTransform.SetCheckMeanDistance(int(checkMeanDistance))
icpTransform.SetMaximumMeanDistance(maximumMeanDistance)
transformFilter = slicer.vtkTransformPolyDataFilter.New()
transformFilter.SetInput(inputSurface.GetPolyData())
transformFilter.SetTransform(icpTransform)
transformFilter.Update()
outputSurface.SetAndObservePolyData(transformFilter.GetOutput())
return
SurfaceToolbox.py
This module provides a bunch of operations that one typically performs on a surface, namely decimation, smoothing, computation of normals and surface cleaning. The filters are executed in a pipe after they are manually enabled with the Enabled checkbox.
XML = """<?xml version="1.0" encoding="utf-8"?>
<executable>
<category>Python Modules</category>
<title>Python Surface Toolbox</title>
<description>
Process a surface using several filters.
</description>
<version>1.0</version>
<documentation-url></documentation-url>
<license></license>
<contributor>Luca Antiga and Daniel Blezek</contributor>
<parameters>
<label>Surface Decimation</label>
<description>Parameters for surface decimation</description>
<boolean>
<name>decimationEnabled</name>
<longflag>decimationEnabled</longflag>
<description>Toggle decimation</description>
<label>Enabled</label>
<default>false</default>
</boolean>
<double>
<name>targetReduction</name>
<longflag>targetReduction</longflag>
<description>Target percent reduction of the number of polygons</description>
<label>Target reduction</label>
<default>0.8</default>
<constraints>
<minimum>0</minimum>
<maximum>1</maximum>
<step>0</step>
</constraints>
</double>
<boolean>
<name>boundaryDeletion</name>
<longflag>boundaryDeletion</longflag>
<description>Toggle deletion of boundary vertices</description>
<label>Boundary vertex deletion</label>
<default>true</default>
</boolean>
</parameters>
<parameters>
<label>Surface Smoothing</label>
<description>Parameters for surface smoothing</description>
<boolean>
<name>smoothingEnabled</name>
<longflag>smoothingEnabled</longflag>
<description>Toggle smoothing</description>
<label>Enabled</label>
<default>false</default>
</boolean>
<string-enumeration>
<name>smoothingMethod</name>
<longflag>smoothingMethod</longflag>
<description>Selection of smoothing method between Laplace and Taubin</description>
<label>Smoothing method</label>
<default>Taubin</default>
<element>Laplace</element>
<element>Taubin</element>
</string-enumeration>
<boolean>
<name>boundarySmoothing</name>
<longflag>boundarySmoothing</longflag>
<description>Toggle allow smoothing of boundary</description>
<label>Boundary smoothing</label>
<default>true</default>
</boolean>
<integer>
<name>laplaceIterations</name>
<longflag>laplaceIterations</longflag>
<description>Number of iterations for the Laplace smoothing method</description>
<label>Laplace number of iterations</label>
<default>10</default>
</integer>
<double>
<name>laplaceRelaxation</name>
<longflag>laplaceRelaxation</longflag>
<description>Relaxation factor value for the Laplace smoothing method</description>
<label>Laplace relaxation factor</label>
<default>0.1</default>
</double>
<integer>
<name>taubinIterations</name>
<longflag>taubinIterations</longflag>
<description>Number of iterations for the Taubin smoothing method</description>
<label>Taubin number of iterations</label>
<default>10</default>
</integer>
<double>
<name>taubinPassband</name>
<longflag>taubinPassband</longflag>
<description>Passband value for the Taubin smoothing method</description>
<label>Taubin passband</label>
<default>0.1</default>
</double>
</parameters>
<parameters>
<label>Surface Normals</label>
<description>Parameters for surface normals</description>
<boolean>
<name>normalsEnabled</name>
<longflag>normalsEnabled</longflag>
<description>Toggle normals</description>
<label>Enabled</label>
<default>false</default>
</boolean>
<boolean>
<name>flipNormals</name>
<longflag>flipNormals</longflag>
<description>Toggle flip normals after computing</description>
<label>Flip normals</label>
<default>false</default>
</boolean>
<boolean>
<name>splitting</name>
<longflag>splitting</longflag>
<description>Toggle splitting surface based on feature angle</description>
<label>Splitting</label>
<default>false</default>
</boolean>
<double>
<name>featureAngle</name>
<longflag>featureAngle</longflag>
<description>Feature angle value for surface splitting</description>
<label>Feature angle</label>
<default>30</default>
<constraints>
<minimum>0</minimum>
<maximum>180</maximum>
<step>1</step>
</constraints>
</double>
</parameters>
<parameters>
<label>Surface Cleaner</label>
<description>Parameters for surface cleaner</description>
<boolean>
<name>cleanerEnabled</name>
<longflag>cleanerEnabled</longflag>
<description>Toggle cleaner</description>
<label>Enabled</label>
<default>false</default>
</boolean>
</parameters>
<parameters>
<label>IO</label>
<description>Input/output parameters</description>
<geometry>
<name>inputSurface</name>
<label>Input Surface</label>
<channel>input</channel>
<index>0</index>
<description>Input surface to be filtered</description>
</geometry>
<geometry>
<name>outputSurface</name>
<label>Output Surface</label>
<channel>output</channel>
<index>1</index>
<description>Output filtered surface</description>
</geometry>
</parameters>
</executable>
"""
def Execute (inputSurface, outputSurface, \
decimationEnabled=False, targetReduction=0.8, boundaryDeletion=True, \
smoothingEnabled=False, smoothingMethod="Taubin", boundarySmoothing=True, laplaceIterations=10, laplaceRelaxation=0.1, taubinIterations=10, taubinPassband=0.1, \
normalsEnabled=False, flipNormals=False, splitting=False, featureAngle=30.0, \
cleanerEnabled=False):
Slicer = __import__("Slicer")
slicer = Slicer.Slicer()
scene = slicer.MRMLScene
inputSurface = scene.GetNodeByID(inputSurface)
outputSurface = scene.GetNodeByID(outputSurface)
surface = inputSurface.GetPolyData()
if decimationEnabled:
decimation = slicer.vtkDecimatePro.New()
decimation.SetInput(surface)
decimation.SetTargetReduction(targetReduction)
if boundaryDeletion:
decimation.BoundaryVertexDeletionOn()
else:
decimation.BoundaryVertexDeletionOff()
decimation.PreserveTopologyOn()
decimation.Update()
surface = decimation.GetOutput()
if smoothingEnabled:
if smoothingMethod == "Laplace":
smoothing = slicer.vtkSmoothPolyDataFilter.New()
smoothing.SetInput(surface)
if boundarySmoothing:
smoothing.BoundarySmoothingOn()
else:
smoothing.BoundarySmoothingOff()
smoothing.SetNumberOfIterations(laplaceIterations)
smoothing.SetRelaxationFactor(laplaceRelaxation)
smoothing.Update()
surface = smoothing.GetOutput()
elif smoothingMethod == "Taubin":
smoothing = slicer.vtkWindowedSincPolyDataFilter.New()
smoothing.SetInput(surface)
if boundarySmoothing:
smoothing.BoundarySmoothingOn()
else:
smoothing.BoundarySmoothingOff()
smoothing.SetNumberOfIterations(taubinIterations)
smoothing.SetPassBand(taubinPassband)
smoothing.Update()
surface = smoothing.GetOutput()
if normalsEnabled:
normals = slicer.vtkPolyDataNormals.New()
normals.SetInput(surface)
normals.AutoOrientNormalsOn()
if flipNormals:
normals.FlipNormalsOn()
else:
normals.FlipNormalsOff()
if splitting:
normals.SplittingOn()
normals.SetFeatureAngle(featureAngle)
else:
normals.SplittingOff()
normals.ConsistencyOn()
normals.Update()
surface = normals.GetOutput()
if cleanerEnabled:
cleaner = slicer.vtkCleanPolyData.New()
cleaner.SetInput(surface)
cleaner.Update()
surface = cleaner.GetOutput()
outputSurface.SetAndObservePolyData(surface)
return
