AbstractScriptedSegmentEditorAutoCompleteEffect.py

Go to the documentation of this file.

import logging
 
import ctk
import qt
import vtk
 
import slicer
from slicer.i18n import tr as _
 
from .AbstractScriptedSegmentEditorEffect import *
 
__all__ = ["AbstractScriptedSegmentEditorAutoCompleteEffect"]
 
 
#
# Abstract class of python scripted segment editor auto-complete effects
#
# Auto-complete effects are a subtype of general effects that allow preview
# and refinement of segmentation results before accepting them.
#
 
 
class AbstractScriptedSegmentEditorAutoCompleteEffect(AbstractScriptedSegmentEditorEffect):
    """AutoCompleteEffect is an effect that can create a full segmentation
    from a partial segmentation (not all slices are segmented or only
    part of the target structures are painted).
    """
 
    def __init__(self, scriptedEffect):
        # Indicates that effect does not operate on one segment, but the whole segmentation.
        # This means that while this effect is active, no segment can be selected
        scriptedEffect.perSegment = False
        AbstractScriptedSegmentEditorEffect.__init__(self, scriptedEffect)
 
        # Number of segments required when editable area is not specified
        self.minimumNumberOfSegments = 1
        # Number of segments required when editable area is specified
        self.minimumNumberOfSegmentsWithEditableArea = 1
        self.clippedMasterImageDataRequired = False
        self.clippedMaskImageDataRequired = False
 
        # Stores merged labelmap image geometry (voxel data is not allocated)
        self.mergedLabelmapGeometryImage = None
        self.selectedSegmentIds = None
        self.selectedSegmentModifiedTimes = {}  # map from segment ID to ModifiedTime
        self.clippedMasterImageData = None
        self.clippedMaskImageData = None
 
        # Observation for auto-update
        self.observedSegmentation = None
        self.segmentationNodeObserverTags = []
 
        # Wait this much after the last modified event before starting aut-update:
        autoUpdateDelaySec = 1.0
        self.delayedAutoUpdateTimer = qt.QTimer()
        self.delayedAutoUpdateTimer.setSingleShot(True)
        self.delayedAutoUpdateTimer.interval = autoUpdateDelaySec * 1000
        self.delayedAutoUpdateTimer.connect("timeout()", self.onPreview)
 
        self.extentGrowthRatio = 0.1  # extent of seed region will be grown outside by this much
        self.minimumExtentMargin = 3
 
        self.previewComputationInProgress = False
 
    def __del__(self):
        self.delayedAutoUpdateTimer.stop()
        self.observeSegmentation(False)
 
    def cleanup(self):
        # Disconnect the timer signal to allow proper garbage collection.
        #
        # This prevents lingering signal/slot connections from keeping
        # the object alive. For more details, see the parent class
        # cleanup() docstring or the following issue:
        # https://github.com/Slicer/Slicer/issues/7392
        self.delayedAutoUpdateTimer.disconnect("timeout()", self.onPreview)
 
    @staticmethod
    def isBackgroundLabelmap(labelmapOrientedImageData, label=None):
        if labelmapOrientedImageData is None:
            return False
        # If five or more corner voxels of the image contain non-zero, then it is background
        extent = labelmapOrientedImageData.GetExtent()
        if extent[0] > extent[1] or extent[2] > extent[3] or extent[4] > extent[5]:
            return False
        numberOfFilledCorners = 0
        for i in [0, 1]:
            for j in [2, 3]:
                for k in [4, 5]:
                    voxelValue = labelmapOrientedImageData.GetScalarComponentAsFloat(extent[i], extent[j], extent[k], 0)
                    if label is None:
                        if voxelValue > 0:
                            numberOfFilledCorners += 1
                    else:
                        if voxelValue == label:
                            numberOfFilledCorners += 1
                    if numberOfFilledCorners > 4:
                        return True
        return False
 
    def setupOptionsFrame(self):
        self.autoUpdateCheckBox = qt.QCheckBox(_("Auto-update"))
        self.autoUpdateCheckBox.setToolTip(_("Auto-update results preview when input segments change."))
        self.autoUpdateCheckBox.setChecked(True)
        self.autoUpdateCheckBox.setEnabled(False)
 
        self.previewButton = qt.QPushButton(_("Initialize"))
        self.previewButton.objectName = self.__class__.__name__ + "Preview"
        self.previewButton.setToolTip(_("Preview complete segmentation"))
        # qt.QSizePolicy(qt.QSizePolicy.Expanding, qt.QSizePolicy.Expanding)
        # fails on some systems, therefore set the policies using separate method calls
        qSize = qt.QSizePolicy()
        qSize.setHorizontalPolicy(qt.QSizePolicy.Expanding)
        self.previewButton.setSizePolicy(qSize)
 
        previewFrame = qt.QHBoxLayout()
        previewFrame.addWidget(self.autoUpdateCheckBox)
        previewFrame.addWidget(self.previewButton)
        self.scriptedEffect.addLabeledOptionsWidget(_("Preview:"), previewFrame)
 
        self.previewOpacitySlider = ctk.ctkSliderWidget()
        self.previewOpacitySlider.setToolTip(_("Adjust visibility of results preview."))
        self.previewOpacitySlider.minimum = 0
        self.previewOpacitySlider.maximum = 1.0
        self.previewOpacitySlider.value = 0.0
        self.previewOpacitySlider.singleStep = 0.05
        self.previewOpacitySlider.pageStep = 0.1
        self.previewOpacitySlider.spinBoxVisible = False
 
        self.previewShow3DButton = qt.QPushButton(_("Show 3D"))
        self.previewShow3DButton.setToolTip(_("Preview results in 3D."))
        self.previewShow3DButton.setCheckable(True)
 
        displayFrame = qt.QHBoxLayout()
        displayFrame.addWidget(qt.QLabel(_("inputs")))
        displayFrame.addWidget(self.previewOpacitySlider)
        displayFrame.addWidget(qt.QLabel(_("results")))
        displayFrame.addWidget(self.previewShow3DButton)
        self.scriptedEffect.addLabeledOptionsWidget(_("Display:"), displayFrame)
 
        self.cancelButton = qt.QPushButton(_("Cancel"))
        self.cancelButton.objectName = self.__class__.__name__ + "Cancel"
        self.cancelButton.setToolTip(_("Clear preview and cancel auto-complete"))
 
        self.applyButton = qt.QPushButton(_("Apply"))
        self.applyButton.objectName = self.__class__.__name__ + "Apply"
        self.applyButton.setToolTip(_("Replace segments by previewed result"))
 
        finishFrame = qt.QHBoxLayout()
        finishFrame.addWidget(self.cancelButton)
        finishFrame.addWidget(self.applyButton)
        self.scriptedEffect.addOptionsWidget(finishFrame)
 
        self.previewButton.connect("clicked()", self.onPreview)
        self.cancelButton.connect("clicked()", self.onCancel)
        self.applyButton.connect("clicked()", self.onApply)
        self.previewOpacitySlider.connect("valueChanged(double)", self.updateMRMLFromGUI)
        self.previewShow3DButton.connect("toggled(bool)", self.updateMRMLFromGUI)
        self.autoUpdateCheckBox.connect("stateChanged(int)", self.updateMRMLFromGUI)
 
    def createCursor(self, widget):
        # Turn off effect-specific cursor for this effect
        return slicer.util.mainWindow().cursor
 
    def setMRMLDefaults(self):
        self.scriptedEffect.setParameterDefault("AutoUpdate", "1")
 
    def onSegmentationModified(self, caller, event):
        if not self.autoUpdateCheckBox.isChecked():
            # just in case a queued request comes through
            return
 
        import vtkSegmentationCorePython as vtkSegmentationCore
 
        segmentationNode = self.scriptedEffect.parameterSetNode().GetSegmentationNode()
        segmentation = segmentationNode.GetSegmentation()
 
        updateNeeded = False
        for segmentIndex in range(self.selectedSegmentIds.GetNumberOfValues()):
            segmentID = self.selectedSegmentIds.GetValue(segmentIndex)
            segment = segmentation.GetSegment(segmentID)
            if not segment:
                # selected segment was deleted, cancel segmentation
                logging.debug("Segmentation operation is cancelled because an input segment was deleted")
                self.onCancel()
                slicer.util.showStatusMessage(_("Segmentation operation is cancelled because an input segment was deleted."), 3000)
                return
            segmentLabelmap = segment.GetRepresentation(vtkSegmentationCore.vtkSegmentationConverter.GetSegmentationBinaryLabelmapRepresentationName())
            if segmentID in self.selectedSegmentModifiedTimes \
                    and segmentLabelmap and segmentLabelmap.GetMTime() == self.selectedSegmentModifiedTimes[segmentID]:
                # this segment has not changed since last update
                continue
            if segmentLabelmap:
                self.selectedSegmentModifiedTimes[segmentID] = segmentLabelmap.GetMTime()
            elif segmentID in self.selectedSegmentModifiedTimes:
                self.selectedSegmentModifiedTimes.pop(segmentID)
            updateNeeded = True
            # continue so that all segment modified times are updated
 
        if not updateNeeded:
            return
 
        logging.debug("Segmentation update requested")
        # There could be multiple update events for a single paint operation (e.g., one segment overwrites the other)
        # therefore don't update directly, just set up/reset a timer that will perform the update when it elapses.
        if not self.previewComputationInProgress:
            self.delayedAutoUpdateTimer.start()
 
    def observeSegmentation(self, observationEnabled):
        if not self.scriptedEffect:
            # This can be the case if the Python interpreter has already begun its shutdown process
            # when observeSegmentation is called from __del__ method.
            return
 
        import vtkSegmentationCorePython as vtkSegmentationCore
 
        parameterSetNode = self.scriptedEffect.parameterSetNode()
        segmentationNode = None
        if parameterSetNode:
            segmentationNode = parameterSetNode.GetSegmentationNode()
 
        segmentation = None
        if segmentationNode:
            segmentation = segmentationNode.GetSegmentation()
 
        if observationEnabled and self.observedSegmentation == segmentation:
            return
        if not observationEnabled and not self.observedSegmentation:
            return
        # Need to update the observer
        # Remove old observer
        if self.observedSegmentation:
            for tag in self.segmentationNodeObserverTags:
                self.observedSegmentation.RemoveObserver(tag)
            self.segmentationNodeObserverTags = []
            self.observedSegmentation = None
        # Add new observer
        if observationEnabled and segmentation is not None:
            self.observedSegmentation = segmentation
            observedEvents = [
                vtkSegmentationCore.vtkSegmentation.SegmentAdded,
                vtkSegmentationCore.vtkSegmentation.SegmentRemoved,
                vtkSegmentationCore.vtkSegmentation.SegmentModified,
                vtkSegmentationCore.vtkSegmentation.SourceRepresentationModified]
            for eventId in observedEvents:
                self.segmentationNodeObserverTags.append(self.observedSegmentation.AddObserver(eventId, self.onSegmentationModified))
 
    def getPreviewNode(self):
        previewNode = self.scriptedEffect.nodeReference(ResultPreviewNodeReferenceRole)
        if (previewNode
                and self.scriptedEffect.parameterDefined("SegmentationResultPreviewOwnerEffect")
                and self.scriptedEffect.parameter("SegmentationResultPreviewOwnerEffect") != self.scriptedEffect.name):
            # another effect owns this preview node
            return None
        return previewNode
 
    def updateGUIFromMRML(self):
        previewNode = self.getPreviewNode()
 
        self.cancelButton.setEnabled(previewNode is not None)
        self.applyButton.setEnabled(previewNode is not None)
 
        self.previewOpacitySlider.setEnabled(previewNode is not None)
        if previewNode:
            wasBlocked = self.previewOpacitySlider.blockSignals(True)
            self.previewOpacitySlider.value = self.getPreviewOpacity()
            self.previewOpacitySlider.blockSignals(wasBlocked)
            self.previewButton.text = _("Update")
            self.previewShow3DButton.setEnabled(True)
            self.previewShow3DButton.setChecked(self.getPreviewShow3D())
            self.autoUpdateCheckBox.setEnabled(True)
            self.observeSegmentation(self.autoUpdateCheckBox.isChecked())
        else:
            self.previewButton.text = _("Initialize")
            self.autoUpdateCheckBox.setEnabled(False)
            self.previewShow3DButton.setEnabled(False)
            self.delayedAutoUpdateTimer.stop()
            self.observeSegmentation(False)
 
        autoUpdate = qt.Qt.Unchecked if self.scriptedEffect.integerParameter("AutoUpdate") == 0 else qt.Qt.Checked
        wasBlocked = self.autoUpdateCheckBox.blockSignals(True)
        self.autoUpdateCheckBox.setCheckState(autoUpdate)
        self.autoUpdateCheckBox.blockSignals(wasBlocked)
 
    def updateMRMLFromGUI(self):
        segmentationNode = self.scriptedEffect.parameterSetNode().GetSegmentationNode()
        previewNode = self.getPreviewNode()
        if previewNode:
            self.setPreviewOpacity(self.previewOpacitySlider.value)
            self.setPreviewShow3D(self.previewShow3DButton.checked)
 
        autoUpdate = 1 if self.autoUpdateCheckBox.isChecked() else 0
        self.scriptedEffect.setParameter("AutoUpdate", autoUpdate)
 
    def onPreview(self):
        if self.previewComputationInProgress:
            return
        self.previewComputationInProgress = True
 
        slicer.util.showStatusMessage(_("Running {effectName} auto-complete...").format(effectName=self.scriptedEffect.name), 2000)
        try:
            with slicer.util.tryWithErrorDisplay(_("Segmentation operation failed:"), waitCursor=True):
                self.preview()
        finally:
            self.previewComputationInProgress = False
 
    def reset(self):
        self.delayedAutoUpdateTimer.stop()
        self.observeSegmentation(False)
        previewNode = self.scriptedEffect.nodeReference(ResultPreviewNodeReferenceRole)
        if previewNode:
            self.scriptedEffect.setCommonNodeReference(ResultPreviewNodeReferenceRole, None)
            slicer.mrmlScene.RemoveNode(previewNode)
            self.scriptedEffect.setCommonParameter("SegmentationResultPreviewOwnerEffect", "")
        segmentationNode = self.scriptedEffect.parameterSetNode().GetSegmentationNode()
        segmentationNode.GetDisplayNode().SetOpacity(1.0)
        self.mergedLabelmapGeometryImage = None
        self.selectedSegmentIds = None
        self.selectedSegmentModifiedTimes = {}
        self.clippedMasterImageData = None
        self.clippedMaskImageData = None
        self.updateGUIFromMRML()
 
    def onCancel(self):
        self.reset()
 
    def onApply(self):
        self.delayedAutoUpdateTimer.stop()
        self.observeSegmentation(False)
 
        segmentationNode = self.scriptedEffect.parameterSetNode().GetSegmentationNode()
        segmentationDisplayNode = segmentationNode.GetDisplayNode()
        previewNode = self.getPreviewNode()
 
        self.scriptedEffect.saveStateForUndo()
 
        previewContainsClosedSurfaceRepresentation = previewNode.GetSegmentation().ContainsRepresentation(
            slicer.vtkSegmentationConverter.GetSegmentationClosedSurfaceRepresentationName())
 
        # Move segments from preview into current segmentation
        segmentIDs = vtk.vtkStringArray()
        previewNode.GetSegmentation().GetSegmentIDs(segmentIDs)
        for index in range(segmentIDs.GetNumberOfValues()):
            segmentID = segmentIDs.GetValue(index)
            previewSegmentLabelmap = slicer.vtkOrientedImageData()
            previewNode.GetBinaryLabelmapRepresentation(segmentID, previewSegmentLabelmap)
            self.scriptedEffect.modifySegmentByLabelmap(segmentationNode, segmentID, previewSegmentLabelmap,
                                                        slicer.qSlicerSegmentEditorAbstractEffect.ModificationModeSet)
            if segmentationDisplayNode is not None and self.isBackgroundLabelmap(previewSegmentLabelmap):
                # Automatically hide result segments that are background (all eight corners are non-zero)
                segmentationDisplayNode.SetSegmentVisibility(segmentID, False)
            previewNode.GetSegmentation().RemoveSegment(segmentID)  # delete now to limit memory usage
 
        if previewContainsClosedSurfaceRepresentation:
            segmentationNode.CreateClosedSurfaceRepresentation()
 
        self.reset()
 
    def setPreviewOpacity(self, opacity):
        segmentationNode = self.scriptedEffect.parameterSetNode().GetSegmentationNode()
        segmentationNode.GetDisplayNode().SetOpacity(1.0 - opacity)
        previewNode = self.getPreviewNode()
        if previewNode:
            previewNode.GetDisplayNode().SetOpacity(opacity)
            previewNode.GetDisplayNode().SetOpacity3D(opacity)
 
        # Make sure the GUI is up-to-date
        wasBlocked = self.previewOpacitySlider.blockSignals(True)
        self.previewOpacitySlider.value = opacity
        self.previewOpacitySlider.blockSignals(wasBlocked)
 
    def getPreviewOpacity(self):
        previewNode = self.getPreviewNode()
        return previewNode.GetDisplayNode().GetOpacity() if previewNode else 0.6  # default opacity for preview
 
    def setPreviewShow3D(self, show):
        previewNode = self.getPreviewNode()
        if previewNode:
            if show:
                previewNode.CreateClosedSurfaceRepresentation()
            else:
                previewNode.RemoveClosedSurfaceRepresentation()
 
        # Make sure the GUI is up-to-date
        wasBlocked = self.previewShow3DButton.blockSignals(True)
        self.previewShow3DButton.checked = show
        self.previewShow3DButton.blockSignals(wasBlocked)
 
    def getPreviewShow3D(self):
        previewNode = self.getPreviewNode()
        if not previewNode:
            return False
        containsClosedSurfaceRepresentation = previewNode.GetSegmentation().ContainsRepresentation(
            slicer.vtkSegmentationConverter.GetSegmentationClosedSurfaceRepresentationName())
        return containsClosedSurfaceRepresentation
 
    def effectiveExtentChanged(self):
        if self.getPreviewNode() is None:
            return True
        if self.mergedLabelmapGeometryImage is None:
            return True
        if self.selectedSegmentIds is None:
            return True
 
        import vtkSegmentationCorePython as vtkSegmentationCore
 
        segmentationNode = self.scriptedEffect.parameterSetNode().GetSegmentationNode()
 
        # The effective extent for the current input segments
        effectiveGeometryImage = slicer.vtkOrientedImageData()
        effectiveGeometryString = segmentationNode.GetSegmentation().DetermineCommonLabelmapGeometry(
            vtkSegmentationCore.vtkSegmentation.EXTENT_UNION_OF_EFFECTIVE_SEGMENTS, self.selectedSegmentIds)
        if not effectiveGeometryString:
            return True
        vtkSegmentationCore.vtkSegmentationConverter.DeserializeImageGeometry(effectiveGeometryString, effectiveGeometryImage)
 
        masterImageData = self.scriptedEffect.sourceVolumeImageData()
        masterImageExtent = masterImageData.GetExtent()
 
        # The effective extent of the selected segments
        effectiveLabelExtent = effectiveGeometryImage.GetExtent()
        # Current extent used for auto-complete preview
        currentLabelExtent = self.mergedLabelmapGeometryImage.GetExtent()
 
        # Determine if the current merged labelmap extent has less than a 3 voxel margin around the effective segment extent
        # (limited by the source image extent)
        return ((masterImageExtent[0] != currentLabelExtent[0] and currentLabelExtent[0] > effectiveLabelExtent[0] - self.minimumExtentMargin) or
                (masterImageExtent[1] != currentLabelExtent[1] and currentLabelExtent[1] < effectiveLabelExtent[1] + self.minimumExtentMargin) or
                (masterImageExtent[2] != currentLabelExtent[2] and currentLabelExtent[2] > effectiveLabelExtent[2] - self.minimumExtentMargin) or
                (masterImageExtent[3] != currentLabelExtent[3] and currentLabelExtent[3] < effectiveLabelExtent[3] + self.minimumExtentMargin) or
                (masterImageExtent[4] != currentLabelExtent[4] and currentLabelExtent[4] > effectiveLabelExtent[4] - self.minimumExtentMargin) or
                (masterImageExtent[5] != currentLabelExtent[5] and currentLabelExtent[5] < effectiveLabelExtent[5] + self.minimumExtentMargin))
 
    def preview(self):
        # Get source volume image data
        import vtkSegmentationCorePython as vtkSegmentationCore
 
        # Get segmentation
        segmentationNode = self.scriptedEffect.parameterSetNode().GetSegmentationNode()
 
        previewNode = self.getPreviewNode()
        previewOpacity = self.getPreviewOpacity()
        previewShow3D = self.getPreviewShow3D()
 
        # If the selectedSegmentIds have been specified, then they shouldn't be overwritten here
        currentSelectedSegmentIds = self.selectedSegmentIds
 
        if self.effectiveExtentChanged():
            self.reset()
 
            # Restore the selectedSegmentIds
            self.selectedSegmentIds = currentSelectedSegmentIds
            if self.selectedSegmentIds is None:
                self.selectedSegmentIds = vtk.vtkStringArray()
                segmentationNode.GetDisplayNode().GetVisibleSegmentIDs(self.selectedSegmentIds)
 
            if self.minimumNumberOfSegments != self.minimumNumberOfSegmentsWithEditableArea:
                editableAreaSpecified = (
                    self.scriptedEffect.parameterSetNode().GetSourceVolumeIntensityMask()
                    or self.scriptedEffect.parameterSetNode().GetMaskMode() != slicer.vtkMRMLSegmentationNode.EditAllowedEverywhere)
                if editableAreaSpecified and self.selectedSegmentIds.GetNumberOfValues() < self.minimumNumberOfSegmentsWithEditableArea:
                    logging.error(f"Auto-complete operation failed: at least {self.minimumNumberOfSegmentsWithEditableArea} visible segments are required when editable area is defined")
                    raise RuntimeError(
                        _("Minimum {minimumNumberOfSegments} visible segments are required.").format(
                            minimumNumberOfSegments=self.minimumNumberOfSegmentsWithEditableArea))
                elif (not editableAreaSpecified) and self.selectedSegmentIds.GetNumberOfValues() < self.minimumNumberOfSegments:
                    logging.error(f"Auto-complete operation skipped: at least {self.minimumNumberOfSegmentsWithEditableArea} visible segments or setting of editable area is required")
                    raise RuntimeError(
                        _("Minimum {minimumNumberOfSegments} visible segments (or specification of editable area or intensity range) is required.").format(
                            minimumNumberOfSegments=self.minimumNumberOfSegments))
            elif self.selectedSegmentIds.GetNumberOfValues() < self.minimumNumberOfSegments:
                # Same number of input segments required regardless of editable area
                logging.error(f"Auto-complete operation failed: at least {self.minimumNumberOfSegments} visible segments are required")
                self.selectedSegmentIds = None
                raise RuntimeError(
                    _("Minimum {minimumNumberOfSegments} visible segments are required.").format(
                        minimumNumberOfSegments=self.minimumNumberOfSegments))
 
            # Compute merged labelmap extent (effective extent slightly expanded)
            if not self.mergedLabelmapGeometryImage:
                self.mergedLabelmapGeometryImage = slicer.vtkOrientedImageData()
            commonGeometryString = segmentationNode.GetSegmentation().DetermineCommonLabelmapGeometry(
                vtkSegmentationCore.vtkSegmentation.EXTENT_UNION_OF_EFFECTIVE_SEGMENTS, self.selectedSegmentIds)
            if not commonGeometryString:
                logging.info("Auto-complete operation skipped: all visible segments are empty")
                return
            vtkSegmentationCore.vtkSegmentationConverter.DeserializeImageGeometry(commonGeometryString, self.mergedLabelmapGeometryImage)
 
            masterImageData = self.scriptedEffect.sourceVolumeImageData()
            masterImageExtent = masterImageData.GetExtent()
            labelsEffectiveExtent = self.mergedLabelmapGeometryImage.GetExtent()
            # Margin size is relative to combined seed region size, but minimum of 3 voxels
            print(f"self.extentGrowthRatio = {self.extentGrowthRatio}")
            margin = [
                int(max(3, self.extentGrowthRatio * (labelsEffectiveExtent[1] - labelsEffectiveExtent[0]))),
                int(max(3, self.extentGrowthRatio * (labelsEffectiveExtent[3] - labelsEffectiveExtent[2]))),
                int(max(3, self.extentGrowthRatio * (labelsEffectiveExtent[5] - labelsEffectiveExtent[4])))]
            labelsExpandedExtent = [
                max(masterImageExtent[0], labelsEffectiveExtent[0] - margin[0]),
                min(masterImageExtent[1], labelsEffectiveExtent[1] + margin[0]),
                max(masterImageExtent[2], labelsEffectiveExtent[2] - margin[1]),
                min(masterImageExtent[3], labelsEffectiveExtent[3] + margin[1]),
                max(masterImageExtent[4], labelsEffectiveExtent[4] - margin[2]),
                min(masterImageExtent[5], labelsEffectiveExtent[5] + margin[2])]
            print("masterImageExtent = " + repr(masterImageExtent))
            print("labelsEffectiveExtent = " + repr(labelsEffectiveExtent))
            print("labelsExpandedExtent = " + repr(labelsExpandedExtent))
            self.mergedLabelmapGeometryImage.SetExtent(labelsExpandedExtent)
 
            # Create and setup preview node
            previewNode = slicer.mrmlScene.AddNewNodeByClass("vtkMRMLSegmentationNode")
            previewNode.CreateDefaultDisplayNodes()
            previewNode.GetDisplayNode().SetVisibility2DOutline(False)
            if segmentationNode.GetParentTransformNode():
                previewNode.SetAndObserveTransformNodeID(segmentationNode.GetParentTransformNode().GetID())
            self.scriptedEffect.setCommonNodeReference(ResultPreviewNodeReferenceRole, previewNode)
            self.scriptedEffect.setCommonParameter("SegmentationResultPreviewOwnerEffect", self.scriptedEffect.name)
            self.setPreviewOpacity(0.6)
 
            # Disable smoothing for closed surface generation to make it fast
            previewNode.GetSegmentation().SetConversionParameter(
                slicer.vtkBinaryLabelmapToClosedSurfaceConversionRule.GetSmoothingFactorParameterName(),
                "-0.5")
 
            inputContainsClosedSurfaceRepresentation = segmentationNode.GetSegmentation().ContainsRepresentation(
                slicer.vtkSegmentationConverter.GetSegmentationClosedSurfaceRepresentationName())
 
            self.setPreviewShow3D(inputContainsClosedSurfaceRepresentation)
 
            if self.clippedMasterImageDataRequired:
                self.clippedMasterImageData = slicer.vtkOrientedImageData()
                masterImageClipper = vtk.vtkImageConstantPad()
                masterImageClipper.SetInputData(masterImageData)
                masterImageClipper.SetOutputWholeExtent(self.mergedLabelmapGeometryImage.GetExtent())
                masterImageClipper.Update()
                self.clippedMasterImageData.ShallowCopy(masterImageClipper.GetOutput())
                self.clippedMasterImageData.CopyDirections(self.mergedLabelmapGeometryImage)
 
            self.clippedMaskImageData = None
            if self.clippedMaskImageDataRequired:
                self.clippedMaskImageData = slicer.vtkOrientedImageData()
                intensityBasedMasking = self.scriptedEffect.parameterSetNode().GetSourceVolumeIntensityMask()
                maskSegmentID = self.scriptedEffect.parameterSetNode().GetMaskSegmentID() if self.scriptedEffect.parameterSetNode().GetMaskSegmentID() else ""
                intensityRange = self.scriptedEffect.parameterSetNode().GetSourceVolumeIntensityMaskRange() if intensityBasedMasking else None
                success = segmentationNode.GenerateEditMask(self.clippedMaskImageData,
                                                            self.scriptedEffect.parameterSetNode().GetMaskMode(),
                                                            self.clippedMasterImageData,  # reference geometry
                                                            "",  # edited segment ID
                                                            maskSegmentID,
                                                            self.clippedMasterImageData if intensityBasedMasking else None,
                                                            intensityRange)
                if not success:
                    logging.error("Failed to create edit mask")
                    self.clippedMaskImageData = None
 
        previewNode.SetName(segmentationNode.GetName() + " preview")
        previewNode.RemoveClosedSurfaceRepresentation()  # Force the closed surface representation to update
        # TODO: This will no longer be required when we can use the segment editor to set multiple segments
        # as the closed surfaces will be converted as necessary by the segmentation logic.
 
        mergedImage = slicer.vtkOrientedImageData()
        segmentationNode.GenerateMergedLabelmapForAllSegments(
            mergedImage,
            vtkSegmentationCore.vtkSegmentation.EXTENT_UNION_OF_EFFECTIVE_SEGMENTS, self.mergedLabelmapGeometryImage, self.selectedSegmentIds)
 
        outputLabelmap = slicer.vtkOrientedImageData()
        self.computePreviewLabelmap(mergedImage, outputLabelmap)
 
        if previewNode.GetSegmentation().GetNumberOfSegments() != self.selectedSegmentIds.GetNumberOfValues():
            # first update (or number of segments changed), need a full reinitialization
            previewNode.GetSegmentation().RemoveAllSegments()
 
        for index in range(self.selectedSegmentIds.GetNumberOfValues()):
            segmentID = self.selectedSegmentIds.GetValue(index)
 
            previewSegment = previewNode.GetSegmentation().GetSegment(segmentID)
            if not previewSegment:
                inputSegment = segmentationNode.GetSegmentation().GetSegment(segmentID)
 
                previewSegment = vtkSegmentationCore.vtkSegment()
                previewSegment.SetName(inputSegment.GetName())
                previewSegment.SetColor(inputSegment.GetColor())
                previewNode.GetSegmentation().AddSegment(previewSegment, segmentID)
 
            labelValue = index + 1  # n-th segment label value = n + 1 (background label value is 0)
            previewSegment.AddRepresentation(vtkSegmentationCore.vtkSegmentationConverter.GetSegmentationBinaryLabelmapRepresentationName(), outputLabelmap)
            previewSegment.SetLabelValue(labelValue)
 
            # Automatically hide result segments that are background (all eight corners are non-zero)
            previewNode.GetDisplayNode().SetSegmentVisibility3D(segmentID, not self.isBackgroundLabelmap(outputLabelmap, labelValue))
 
        # If the preview was reset, we need to restore the visibility options
        self.setPreviewOpacity(previewOpacity)
        self.setPreviewShow3D(previewShow3D)
 
        self.updateGUIFromMRML()
 
 
ResultPreviewNodeReferenceRole = "SegmentationResultPreview"