Segment With Geodesic Active Contour Level Set#

Synopsis#

Segments structures in images based on a user supplied edge potential map.

This example is the same as the one in the ITK software guide. It was re-organized and does not contain any comments about the executed filters. Please refer to the guide for more details.

Results#

Evolution of the segmentation depending on the number of iterations (10 to 500).

Used parameters:

Left ventricle: 81 114 5.0 1.0 -0.5 3.0 2.0
Right ventricle: 99 114 5.0 1.0 -0.5 3.0 2.0
White matter: 56 92 5.0 1.0 -0.3 2.0 10.0
Gray matter: 40 90 5.0 .5 -0.3 2.0 10.0

Code#

Python#

#!/usr/bin/env python

import itk
import argparse

parser = argparse.ArgumentParser(
    description="Segment With Geodesic Active Contour Level Set."
)
parser.add_argument("input_image")
parser.add_argument("output_image")
parser.add_argument("seed_x", type=int)
parser.add_argument("seed_y", type=int)
parser.add_argument("initial_distance", type=float)
parser.add_argument("sigma", type=float)
parser.add_argument("sigmoid_alpha", type=float)
parser.add_argument("sigmoid_beta", type=float)
parser.add_argument("propagation_scaling", type=float)
parser.add_argument("number_of_iterations", type=int)
args = parser.parse_args()

seedValue = -args.initial_distance

Dimension = 2

InputPixelType = itk.F
OutputPixelType = itk.UC

InputImageType = itk.Image[InputPixelType, Dimension]
OutputImageType = itk.Image[OutputPixelType, Dimension]

ReaderType = itk.ImageFileReader[InputImageType]
WriterType = itk.ImageFileWriter[OutputImageType]

reader = ReaderType.New()
reader.SetFileName(args.input_image)

SmoothingFilterType = itk.CurvatureAnisotropicDiffusionImageFilter[
    InputImageType, InputImageType
]
smoothing = SmoothingFilterType.New()
smoothing.SetTimeStep(0.125)
smoothing.SetNumberOfIterations(5)
smoothing.SetConductanceParameter(9.0)
smoothing.SetInput(reader.GetOutput())

GradientFilterType = itk.GradientMagnitudeRecursiveGaussianImageFilter[
    InputImageType, InputImageType
]
gradientMagnitude = GradientFilterType.New()
gradientMagnitude.SetSigma(args.sigma)
gradientMagnitude.SetInput(smoothing.GetOutput())

SigmoidFilterType = itk.SigmoidImageFilter[InputImageType, InputImageType]
sigmoid = SigmoidFilterType.New()
sigmoid.SetOutputMinimum(0.0)
sigmoid.SetOutputMaximum(1.0)
sigmoid.SetAlpha(args.sigmoid_alpha)
sigmoid.SetBeta(args.sigmoid_beta)
sigmoid.SetInput(gradientMagnitude.GetOutput())

FastMarchingFilterType = itk.FastMarchingImageFilter[InputImageType, InputImageType]
fastMarching = FastMarchingFilterType.New()

GeoActiveContourFilterType = itk.GeodesicActiveContourLevelSetImageFilter[
    InputImageType, InputImageType, InputPixelType
]
geodesicActiveContour = GeoActiveContourFilterType.New()
geodesicActiveContour.SetPropagationScaling(args.propagation_scaling)
geodesicActiveContour.SetCurvatureScaling(1.0)
geodesicActiveContour.SetAdvectionScaling(1.0)
geodesicActiveContour.SetMaximumRMSError(0.02)
geodesicActiveContour.SetNumberOfIterations(args.number_of_iterations)
geodesicActiveContour.SetInput(fastMarching.GetOutput())
geodesicActiveContour.SetFeatureImage(sigmoid.GetOutput())

ThresholdingFilterType = itk.BinaryThresholdImageFilter[InputImageType, OutputImageType]
thresholder = ThresholdingFilterType.New()
thresholder.SetLowerThreshold(-1000.0)
thresholder.SetUpperThreshold(0.0)
thresholder.SetOutsideValue(itk.NumericTraits[OutputPixelType].min())
thresholder.SetInsideValue(itk.NumericTraits[OutputPixelType].max())
thresholder.SetInput(geodesicActiveContour.GetOutput())

seedPosition = itk.Index[Dimension]()
seedPosition[0] = args.seed_x
seedPosition[1] = args.seed_y

node = itk.LevelSetNode[InputPixelType, Dimension]()
node.SetValue(seedValue)
node.SetIndex(seedPosition)

seeds = itk.VectorContainer[itk.UI, itk.LevelSetNode[InputPixelType, Dimension]].New()
seeds.Initialize()
seeds.InsertElement(0, node)

fastMarching.SetTrialPoints(seeds)
fastMarching.SetSpeedConstant(1.0)

CastFilterType = itk.RescaleIntensityImageFilter[InputImageType, OutputImageType]

caster1 = CastFilterType.New()
caster2 = CastFilterType.New()
caster3 = CastFilterType.New()
caster4 = CastFilterType.New()

writer1 = WriterType.New()
writer2 = WriterType.New()
writer3 = WriterType.New()
writer4 = WriterType.New()

caster1.SetInput(smoothing.GetOutput())
writer1.SetInput(caster1.GetOutput())
writer1.SetFileName("GeodesicActiveContourImageFilterOutput1.png")
caster1.SetOutputMinimum(itk.NumericTraits[OutputPixelType].min())
caster1.SetOutputMaximum(itk.NumericTraits[OutputPixelType].max())
writer1.Update()

caster2.SetInput(gradientMagnitude.GetOutput())
writer2.SetInput(caster2.GetOutput())
writer2.SetFileName("GeodesicActiveContourImageFilterOutput2.png")
caster2.SetOutputMinimum(itk.NumericTraits[OutputPixelType].min())
caster2.SetOutputMaximum(itk.NumericTraits[OutputPixelType].max())
writer2.Update()

caster3.SetInput(sigmoid.GetOutput())
writer3.SetInput(caster3.GetOutput())
writer3.SetFileName("GeodesicActiveContourImageFilterOutput3.png")
caster3.SetOutputMinimum(itk.NumericTraits[OutputPixelType].min())
caster3.SetOutputMaximum(itk.NumericTraits[OutputPixelType].max())
writer3.Update()

caster4.SetInput(fastMarching.GetOutput())
writer4.SetInput(caster4.GetOutput())
writer4.SetFileName("GeodesicActiveContourImageFilterOutput4.png")
caster4.SetOutputMinimum(itk.NumericTraits[OutputPixelType].min())
caster4.SetOutputMaximum(itk.NumericTraits[OutputPixelType].max())

fastMarching.SetOutputSize(reader.GetOutput().GetBufferedRegion().GetSize())

writer = WriterType.New()
writer.SetFileName(args.output_image)
writer.SetInput(thresholder.GetOutput())
writer.Update()

print(
    "Max. no. iterations: " + str(geodesicActiveContour.GetNumberOfIterations()) + "\n"
)
print("Max. RMS error: " + str(geodesicActiveContour.GetMaximumRMSError()) + "\n")
print(
    "No. elpased iterations: "
    + str(geodesicActiveContour.GetElapsedIterations())
    + "\n"
)
print("RMS change: " + str(geodesicActiveContour.GetRMSChange()) + "\n")

writer4.Update()

InternalWriterType = itk.ImageFileWriter[InputImageType]

mapWriter = InternalWriterType.New()
mapWriter.SetInput(fastMarching.GetOutput())
mapWriter.SetFileName("GeodesicActiveContourImageFilterOutput4.mha")
mapWriter.Update()

speedWriter = InternalWriterType.New()
speedWriter.SetInput(sigmoid.GetOutput())
speedWriter.SetFileName("GeodesicActiveContourImageFilterOutput3.mha")
speedWriter.Update()

gradientWriter = InternalWriterType.New()
gradientWriter.SetInput(gradientMagnitude.GetOutput())
gradientWriter.SetFileName("GeodesicActiveContourImageFilterOutput2.mha")
gradientWriter.Update()

C++#

#include "itkImageFileReader.h"
#include "itkImageFileWriter.h"
#include "itkGeodesicActiveContourLevelSetImageFilter.h"
#include "itkCurvatureAnisotropicDiffusionImageFilter.h"
#include "itkGradientMagnitudeRecursiveGaussianImageFilter.h"
#include "itkSigmoidImageFilter.h"
#include "itkFastMarchingImageFilter.h"
#include "itkRescaleIntensityImageFilter.h"
#include "itkBinaryThresholdImageFilter.h"

int
main(int argc, char * argv[])
{
  if (argc != 11)
  {
    std::cerr << "Usage: " << argv[0];
    std::cerr << " <InputFileName>  <OutputFileName>";
    std::cerr << " <seedX> <seedY> <InitialDistance>";
    std::cerr << " <Sigma> <SigmoidAlpha> <SigmoidBeta>";
    std::cerr << " <PropagationScaling> <NumberOfIterations>" << std::endl;
    return EXIT_FAILURE;
  }

  const char * inputFileName = argv[1];
  const char * outputFileName = argv[2];
  const int    seedPosX = std::stoi(argv[3]);
  const int    seedPosY = std::stoi(argv[4]);

  const double initialDistance = std::stod(argv[5]);
  const double sigma = std::stod(argv[6]);
  const double alpha = std::stod(argv[7]);
  const double beta = std::stod(argv[8]);
  const double propagationScaling = std::stod(argv[9]);
  const double numberOfIterations = std::stoi(argv[10]);
  const double seedValue = -initialDistance;

  constexpr unsigned int Dimension = 2;

  using InputPixelType = float;
  using InputImageType = itk::Image<InputPixelType, Dimension>;
  using OutputPixelType = unsigned char;
  using OutputImageType = itk::Image<OutputPixelType, Dimension>;

  const auto input = itk::ReadImage<InputImageType>(inputFileName);

  using SmoothingFilterType = itk::CurvatureAnisotropicDiffusionImageFilter<InputImageType, InputImageType>;
  auto smoothing = SmoothingFilterType::New();
  smoothing->SetTimeStep(0.125);
  smoothing->SetNumberOfIterations(5);
  smoothing->SetConductanceParameter(9.0);
  smoothing->SetInput(input);

  using GradientFilterType = itk::GradientMagnitudeRecursiveGaussianImageFilter<InputImageType, InputImageType>;
  auto gradientMagnitude = GradientFilterType::New();
  gradientMagnitude->SetSigma(sigma);
  gradientMagnitude->SetInput(smoothing->GetOutput());

  using SigmoidFilterType = itk::SigmoidImageFilter<InputImageType, InputImageType>;
  auto sigmoid = SigmoidFilterType::New();
  sigmoid->SetOutputMinimum(0.0);
  sigmoid->SetOutputMaximum(1.0);
  sigmoid->SetAlpha(alpha);
  sigmoid->SetBeta(beta);
  sigmoid->SetInput(gradientMagnitude->GetOutput());

  using FastMarchingFilterType = itk::FastMarchingImageFilter<InputImageType, InputImageType>;
  auto fastMarching = FastMarchingFilterType::New();

  using GeodesicActiveContourFilterType = itk::GeodesicActiveContourLevelSetImageFilter<InputImageType, InputImageType>;
  auto geodesicActiveContour = GeodesicActiveContourFilterType::New();
  geodesicActiveContour->SetPropagationScaling(propagationScaling);
  geodesicActiveContour->SetCurvatureScaling(1.0);
  geodesicActiveContour->SetAdvectionScaling(1.0);
  geodesicActiveContour->SetMaximumRMSError(0.02);
  geodesicActiveContour->SetNumberOfIterations(numberOfIterations);
  geodesicActiveContour->SetInput(fastMarching->GetOutput());
  geodesicActiveContour->SetFeatureImage(sigmoid->GetOutput());

  using ThresholdingFilterType = itk::BinaryThresholdImageFilter<InputImageType, OutputImageType>;
  auto thresholder = ThresholdingFilterType::New();
  thresholder->SetLowerThreshold(-1000.0);
  thresholder->SetUpperThreshold(0.0);
  thresholder->SetOutsideValue(itk::NumericTraits<OutputPixelType>::min());
  thresholder->SetInsideValue(itk::NumericTraits<OutputPixelType>::max());
  thresholder->SetInput(geodesicActiveContour->GetOutput());

  using NodeContainer = FastMarchingFilterType::NodeContainer;
  using NodeType = FastMarchingFilterType::NodeType;

  InputImageType::IndexType seedPosition;
  seedPosition[0] = seedPosX;
  seedPosition[1] = seedPosY;

  auto     seeds = NodeContainer::New();
  NodeType node;
  node.SetValue(seedValue);
  node.SetIndex(seedPosition);

  seeds->Initialize();
  seeds->InsertElement(0, node);

  fastMarching->SetTrialPoints(seeds);
  fastMarching->SetSpeedConstant(1.0);

  using CastFilterType = itk::RescaleIntensityImageFilter<InputImageType, OutputImageType>;

  auto caster1 = CastFilterType::New();
  auto caster2 = CastFilterType::New();
  auto caster3 = CastFilterType::New();
  auto caster4 = CastFilterType::New();

  caster1->SetInput(smoothing->GetOutput());
  caster1->SetOutputMinimum(itk::NumericTraits<OutputPixelType>::min());
  caster1->SetOutputMaximum(itk::NumericTraits<OutputPixelType>::max());
  itk::WriteImage(caster1->GetOutput(), "GeodesicActiveContourImageFilterOutput1.png");

  caster2->SetInput(gradientMagnitude->GetOutput());
  caster2->SetOutputMinimum(itk::NumericTraits<OutputPixelType>::min());
  caster2->SetOutputMaximum(itk::NumericTraits<OutputPixelType>::max());
  itk::WriteImage(caster2->GetOutput(), "GeodesicActiveContourImageFilterOutput2.png");

  caster3->SetInput(sigmoid->GetOutput());
  caster3->SetOutputMinimum(itk::NumericTraits<OutputPixelType>::min());
  caster3->SetOutputMaximum(itk::NumericTraits<OutputPixelType>::max());
  itk::WriteImage(caster3->GetOutput(), "GeodesicActiveContourImageFilterOutput3.png");

  caster4->SetInput(fastMarching->GetOutput());
  caster4->SetOutputMinimum(itk::NumericTraits<OutputPixelType>::min());
  caster4->SetOutputMaximum(itk::NumericTraits<OutputPixelType>::max());

  fastMarching->SetOutputDirection(input->GetDirection());
  fastMarching->SetOutputOrigin(input->GetOrigin());
  fastMarching->SetOutputRegion(input->GetBufferedRegion());
  fastMarching->SetOutputSpacing(input->GetSpacing());

  try
  {
    itk::WriteImage(thresholder->GetOutput(), outputFileName);
  }
  catch (const itk::ExceptionObject & error)
  {
    std::cerr << "Error: " << error << std::endl;
    return EXIT_FAILURE;
  }

  std::cout << std::endl;
  std::cout << "Max. no. iterations: " << geodesicActiveContour->GetNumberOfIterations() << std::endl;
  std::cout << "Max. RMS error: " << geodesicActiveContour->GetMaximumRMSError() << std::endl;
  std::cout << std::endl;
  std::cout << "No. elpased iterations: " << geodesicActiveContour->GetElapsedIterations() << std::endl;
  std::cout << "RMS change: " << geodesicActiveContour->GetRMSChange() << std::endl;

  try
  {
    itk::WriteImage(caster4->GetOutput(), "GeodesicActiveContourImageFilterOutput4.png");
  }
  catch (const itk::ExceptionObject & error)
  {
    std::cerr << "Error: " << error << std::endl;
    return EXIT_FAILURE;
  }

  try
  {
    itk::WriteImage(fastMarching->GetOutput(), "GeodesicActiveContourImageFilterOutput4.mha");
  }
  catch (const itk::ExceptionObject & error)
  {
    std::cerr << "Error: " << error << std::endl;
    return EXIT_FAILURE;
  }

  try
  {
    itk::WriteImage(sigmoid->GetOutput(), "GeodesicActiveContourImageFilterOutput3.mha");
  }
  catch (const itk::ExceptionObject & error)
  {
    std::cerr << "Error: " << error << std::endl;
    return EXIT_FAILURE;
  }

  try
  {
    itk::WriteImage(gradientMagnitude->GetOutput(), "GeodesicActiveContourImageFilterOutput2.mha");
  }
  catch (const itk::ExceptionObject & error)
  {
    std::cerr << "Error: " << error << std::endl;
    return EXIT_FAILURE;
  }

  return EXIT_SUCCESS;
}

Classes demonstrated#

template<typename TInputImage, typename TFeatureImage, typename TOutputPixelType = float> class GeodesicActiveContourLevelSetImageFilter : public itk::SegmentationLevelSetImageFilter<TInputImage, TFeatureImage, TOutputPixelType>

Segments structures in images based on a user supplied edge potential map.

$g(I) = 1 / ( 1 + | (\nabla * G)(I)| )$

$g(I) = \exp^{-|(\nabla * G)(I)|}$

IMPORTANT: The SegmentationLevelSetImageFilter class and the GeodesicActiveContourLevelSetFunction class contain additional information necessary to gain full understanding of how to use this filter.
OVERVIEW: This class is a level set method segmentation filter. An initial contour is propagated outwards (or inwards) until it ‘’sticks’’ to the shape boundaries. This is done by using a level set speed function based on a user supplied edge potential map.
INPUTS: This filter requires two inputs. The first input is a initial level set. The initial level set is a real image which contains the initial contour/surface as the zero level set. For example, a signed distance function from the initial contour/surface is typically used. Unlike the simpler ShapeDetectionLevelSetImageFilter the initial contour does not have to lie wholly within the shape to be segmented. The initial contour is allow to overlap the shape boundary. The extra advection term in the update equation behaves like a doublet and attracts the contour to the boundary. This approach for segmentation follows that of Caselles et al (1997).
: The second input is the feature image. For this filter, this is the edge potential map. General characteristics of an edge potential map is that it has values close to zero in regions near the edges and values close to one inside the shape itself. Typically, the edge potential map is compute from the image gradient, for example:

where $I$ is image intensity and $(\nabla * G)$ is the derivative of Gaussian operator.

This implementation allows the user to set the weights between the propagation, advection and curvature term using methods

SetPropagationScaling(), SetAdvectionScaling(), SetCurvatureScaling(). In general, the larger the CurvatureScaling, the smoother the resulting contour. To follow the implementation in Caselles et al paper, set the PropagationScaling to $c$ (the inflation or ballon force) and AdvectionScaling and CurvatureScaling both to 1.0.

: See SegmentationLevelSetImageFilter and SparseFieldLevelSetImageFilter for more information on Inputs.
PARAMETERS: The PropagationScaling parameter can be used to switch from propagation outwards (POSITIVE scaling parameter) versus propagating inwards (NEGATIVE scaling parameter).

OUTPUTS: The filter outputs a single, scalar, real-valued image. Negative values in the output image represent the inside of the segmented region and positive values in the image represent the outside of the segmented region. The zero crossings of the image correspond to the position of the propagating front.
: See SparseFieldLevelSetImageFilter and SegmentationLevelSetImageFilter for more information.
REFERENCES
: ”Geodesic Active Contours”, V. Caselles, R. Kimmel and G. Sapiro. International Journal on Computer Vision, Vol 22, No. 1, pp 61-97, 1997
See: SegmentationLevelSetImageFilter
See: GeodesicActiveContourLevelSetFunction
See: SparseFieldLevelSetImageFilter

See itk::GeodesicActiveContourLevelSetImageFilter for additional documentation.