ITK  5.0.0
Insight Segmentation and Registration Toolkit
Examples/RegistrationITKv3/DeformableRegistration15.cxx
/*=========================================================================
*
* Copyright Insight Software Consortium
*
* Licensed under the Apache License, Version 2.0 (the "License");
* you may not use this file except in compliance with the License.
* You may obtain a copy of the License at
*
* http://www.apache.org/licenses/LICENSE-2.0.txt
*
* Unless required by applicable law or agreed to in writing, software
* distributed under the License is distributed on an "AS IS" BASIS,
* WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
* See the License for the specific language governing permissions and
* limitations under the License.
*
*=========================================================================*/
// Software Guide : BeginLatex
//
// This example illustrates a realistic pipeline for solving a full deformable registration problem.
//
// First the two images are roughly aligned by using a transform
// initialization, then they are registered using a rigid transform, that in
// turn, is used to initialize a registration with an affine transform. The
// transform resulting from the affine registration is used as the bulk
// transform of a BSplineTransform. The deformable registration is
// computed, and finally the resulting transform is used to resample the moving
// image.
//
// Software Guide : EndLatex
// Software Guide : BeginLatex
//
// The following are the most relevant headers to this example.
//
// \index{itk::VersorRigid3DTransform!header}
// \index{itk::AffineTransform!header}
// \index{itk::BSplineTransform!header}
// \index{itk::RegularStepGradientDescentOptimizer!header}
//
// Software Guide : EndLatex
// Software Guide : BeginCodeSnippet
// Software Guide : EndCodeSnippet
// The following section of code implements a Command observer
// used to monitor the evolution of the registration process.
//
#include "itkCommand.h"
class CommandIterationUpdate : public itk::Command
{
public:
using Self = CommandIterationUpdate;
using Superclass = itk::Command;
using Pointer = itk::SmartPointer<Self>;
itkNewMacro( Self );
protected:
CommandIterationUpdate() {};
public:
using OptimizerPointer = const OptimizerType *;
void Execute(itk::Object *caller, const itk::EventObject & event) override
{
Execute( (const itk::Object *)caller, event);
}
void Execute(const itk::Object * object, const itk::EventObject & event) override
{
OptimizerPointer optimizer = static_cast< OptimizerPointer >( object );
if( !(itk::IterationEvent().CheckEvent( &event )) )
{
return;
}
std::cout << optimizer->GetCurrentIteration() << " ";
std::cout << optimizer->GetValue() << " ";
std::cout << std::endl;
}
};
int main( int argc, char *argv[] )
{
if( argc < 4 )
{
std::cerr << "Missing Parameters " << std::endl;
std::cerr << "Usage: " << argv[0];
std::cerr << " fixedImageFile movingImageFile outputImagefile ";
std::cerr << " [differenceOutputfile] [differenceBeforeRegistration] ";
std::cerr << " [deformationField] ";
std::cerr << " [useExplicitPDFderivatives ] [useCachingBSplineWeights ] ";
std::cerr << " [filenameForFinalTransformParameters] ";
std::cerr << " [numberOfGridNodesInsideImageInOneDimensionCoarse] ";
std::cerr << " [numberOfGridNodesInsideImageInOneDimensionFine] ";
std::cerr << " [maximumStepLength] [maximumNumberOfIterations]";
std::cerr << std::endl;
return EXIT_FAILURE;
}
constexpr unsigned int ImageDimension = 3;
using PixelType = signed short;
using MovingImageType = itk::Image< PixelType, ImageDimension >;
const unsigned int SpaceDimension = ImageDimension;
constexpr unsigned int SplineOrder = 3;
using CoordinateRepType = double;
using RigidTransformType = itk::VersorRigid3DTransform< double >;
using DeformableTransformType = itk::BSplineTransform<
CoordinateRepType,
SpaceDimension,
SplineOrder >;
using TransformInitializerType = itk::CenteredTransformInitializer< RigidTransformType,
FixedImageType, MovingImageType >;
FixedImageType,
MovingImageType >;
using InterpolatorType = itk:: LinearInterpolateImageFunction<
MovingImageType,
double >;
using RegistrationType = itk::ImageRegistrationMethod<
FixedImageType,
MovingImageType >;
MetricType::Pointer metric = MetricType::New();
OptimizerType::Pointer optimizer = OptimizerType::New();
InterpolatorType::Pointer interpolator = InterpolatorType::New();
RegistrationType::Pointer registration = RegistrationType::New();
registration->SetMetric( metric );
registration->SetOptimizer( optimizer );
registration->SetInterpolator( interpolator );
// Auxiliary identity transform.
using IdentityTransformType = itk::IdentityTransform<double,SpaceDimension>;
IdentityTransformType::Pointer identityTransform = IdentityTransformType::New();
//
// Read the Fixed and Moving images.
//
using FixedImageReaderType = itk::ImageFileReader< FixedImageType >;
using MovingImageReaderType = itk::ImageFileReader< MovingImageType >;
FixedImageReaderType::Pointer fixedImageReader = FixedImageReaderType::New();
MovingImageReaderType::Pointer movingImageReader = MovingImageReaderType::New();
fixedImageReader->SetFileName( argv[1] );
movingImageReader->SetFileName( argv[2] );
try
{
fixedImageReader->Update();
movingImageReader->Update();
}
catch( itk::ExceptionObject & err )
{
std::cerr << "ExceptionObject caught !" << std::endl;
std::cerr << err << std::endl;
return EXIT_FAILURE;
}
FixedImageType::ConstPointer fixedImage = fixedImageReader->GetOutput();
registration->SetFixedImage( fixedImage );
registration->SetMovingImage( movingImageReader->GetOutput() );
//
// Add a time and memory probes collector for profiling the computation time
// of every stage.
//
//
// Setup the metric parameters
//
metric->SetNumberOfHistogramBins( 50 );
FixedImageType::RegionType fixedRegion = fixedImage->GetBufferedRegion();
const unsigned int numberOfPixels = fixedRegion.GetNumberOfPixels();
metric->ReinitializeSeed( 76926294 );
if( argc > 7 )
{
// Define whether to calculate the metric derivative by explicitly
// computing the derivatives of the joint PDF with respect to the Transform
// parameters, or doing it by progressively accumulating contributions from
// each bin in the joint PDF.
metric->SetUseExplicitPDFDerivatives( std::stoi( argv[7] ) );
}
if( argc > 8 )
{
// Define whether to cache the BSpline weights and indexes corresponding to
// each one of the samples used to compute the metric. Enabling caching will
// make the algorithm run faster but it will have a cost on the amount of memory
// that needs to be allocated. This option is only relevant when using the
// BSplineTransform.
metric->SetUseCachingOfBSplineWeights( std::stoi( argv[8] ) );
}
//
// Initialize a rigid transform by using Image Intensity Moments
//
TransformInitializerType::Pointer initializer = TransformInitializerType::New();
RigidTransformType::Pointer rigidTransform = RigidTransformType::New();
initializer->SetTransform( rigidTransform );
initializer->SetFixedImage( fixedImageReader->GetOutput() );
initializer->SetMovingImage( movingImageReader->GetOutput() );
initializer->MomentsOn();
std::cout << "Starting Rigid Transform Initialization " << std::endl;
memorymeter.Start( "Rigid Initialization" );
chronometer.Start( "Rigid Initialization" );
initializer->InitializeTransform();
chronometer.Stop( "Rigid Initialization" );
memorymeter.Stop( "Rigid Initialization" );
std::cout << "Rigid Transform Initialization completed" << std::endl;
std::cout << std::endl;
registration->SetFixedImageRegion( fixedRegion );
registration->SetInitialTransformParameters( rigidTransform->GetParameters() );
registration->SetTransform( rigidTransform );
//
// Define optimizer normaliztion to compensate for different dynamic range
// of rotations and translations.
//
using OptimizerScalesType = OptimizerType::ScalesType;
OptimizerScalesType optimizerScales( rigidTransform->GetNumberOfParameters() );
const double translationScale = 1.0 / 1000.0;
optimizerScales[0] = 1.0;
optimizerScales[1] = 1.0;
optimizerScales[2] = 1.0;
optimizerScales[3] = translationScale;
optimizerScales[4] = translationScale;
optimizerScales[5] = translationScale;
optimizer->SetScales( optimizerScales );
optimizer->SetMaximumStepLength( 0.2000 );
optimizer->SetMinimumStepLength( 0.0001 );
optimizer->SetNumberOfIterations( 200 );
//
// The rigid transform has 6 parameters we use therefore a few samples to run
// this stage.
//
// Regulating the number of samples in the Metric is equivalent to performing
// multi-resolution registration because it is indeed a sub-sampling of the
// image.
metric->SetNumberOfSpatialSamples( 10000L );
//
// Create the Command observer and register it with the optimizer.
//
CommandIterationUpdate::Pointer observer = CommandIterationUpdate::New();
optimizer->AddObserver( itk::IterationEvent(), observer );
std::cout << "Starting Rigid Registration " << std::endl;
try
{
memorymeter.Start( "Rigid Registration" );
chronometer.Start( "Rigid Registration" );
registration->Update();
chronometer.Stop( "Rigid Registration" );
memorymeter.Stop( "Rigid Registration" );
std::cout << "Optimizer stop condition = "
<< registration->GetOptimizer()->GetStopConditionDescription()
<< std::endl;
}
catch( itk::ExceptionObject & err )
{
std::cerr << "ExceptionObject caught !" << std::endl;
std::cerr << err << std::endl;
return EXIT_FAILURE;
}
std::cout << "Rigid Registration completed" << std::endl;
std::cout << std::endl;
rigidTransform->SetParameters( registration->GetLastTransformParameters() );
//
// Perform Affine Registration
//
AffineTransformType::Pointer affineTransform = AffineTransformType::New();
affineTransform->SetCenter( rigidTransform->GetCenter() );
affineTransform->SetTranslation( rigidTransform->GetTranslation() );
affineTransform->SetMatrix( rigidTransform->GetMatrix() );
registration->SetTransform( affineTransform );
registration->SetInitialTransformParameters( affineTransform->GetParameters() );
optimizerScales = OptimizerScalesType( affineTransform->GetNumberOfParameters() );
optimizerScales[0] = 1.0;
optimizerScales[1] = 1.0;
optimizerScales[2] = 1.0;
optimizerScales[3] = 1.0;
optimizerScales[4] = 1.0;
optimizerScales[5] = 1.0;
optimizerScales[6] = 1.0;
optimizerScales[7] = 1.0;
optimizerScales[8] = 1.0;
optimizerScales[9] = translationScale;
optimizerScales[10] = translationScale;
optimizerScales[11] = translationScale;
optimizer->SetScales( optimizerScales );
optimizer->SetMaximumStepLength( 0.2000 );
optimizer->SetMinimumStepLength( 0.0001 );
optimizer->SetNumberOfIterations( 200 );
//
// The Affine transform has 12 parameters we use therefore a more samples to run
// this stage.
//
// Regulating the number of samples in the Metric is equivalent to performing
// multi-resolution registration because it is indeed a sub-sampling of the
// image.
metric->SetNumberOfSpatialSamples( 50000L );
std::cout << "Starting Affine Registration " << std::endl;
try
{
memorymeter.Start( "Affine Registration" );
chronometer.Start( "Affine Registration" );
registration->Update();
chronometer.Stop( "Affine Registration" );
memorymeter.Stop( "Affine Registration" );
}
catch( itk::ExceptionObject & err )
{
std::cerr << "ExceptionObject caught !" << std::endl;
std::cerr << err << std::endl;
return EXIT_FAILURE;
}
std::cout << "Affine Registration completed" << std::endl;
std::cout << std::endl;
affineTransform->SetParameters( registration->GetLastTransformParameters() );
//
// Perform Deformable Registration
//
DeformableTransformType::Pointer bsplineTransformCoarse = DeformableTransformType::New();
unsigned int numberOfGridNodesInOneDimensionCoarse = 5;
DeformableTransformType::PhysicalDimensionsType fixedPhysicalDimensions;
DeformableTransformType::MeshSizeType meshSize;
DeformableTransformType::OriginType fixedOrigin;
for( unsigned int i=0; i< SpaceDimension; i++ )
{
fixedOrigin[i] = fixedImage->GetOrigin()[i];
fixedPhysicalDimensions[i] = fixedImage->GetSpacing()[i] *
static_cast<double>(
fixedImage->GetLargestPossibleRegion().GetSize()[i] - 1 );
}
meshSize.Fill( numberOfGridNodesInOneDimensionCoarse - SplineOrder );
bsplineTransformCoarse->SetTransformDomainOrigin( fixedOrigin );
bsplineTransformCoarse->SetTransformDomainPhysicalDimensions(
fixedPhysicalDimensions );
bsplineTransformCoarse->SetTransformDomainMeshSize( meshSize );
bsplineTransformCoarse->SetTransformDomainDirection(
fixedImage->GetDirection() );
using ParametersType = DeformableTransformType::ParametersType;
unsigned int numberOfBSplineParameters = bsplineTransformCoarse->GetNumberOfParameters();
optimizerScales = OptimizerScalesType( numberOfBSplineParameters );
optimizerScales.Fill( 1.0 );
optimizer->SetScales( optimizerScales );
ParametersType initialDeformableTransformParameters( numberOfBSplineParameters );
initialDeformableTransformParameters.Fill( 0.0 );
bsplineTransformCoarse->SetParameters( initialDeformableTransformParameters );
registration->SetInitialTransformParameters( bsplineTransformCoarse->GetParameters() );
registration->SetTransform( bsplineTransformCoarse );
// Software Guide : EndCodeSnippet
// Software Guide : BeginLatex
//
// Next we set the parameters of the RegularStepGradientDescentOptimizer object.
//
// Software Guide : EndLatex
// Software Guide : BeginCodeSnippet
optimizer->SetMaximumStepLength( 10.0 );
optimizer->SetMinimumStepLength( 0.01 );
optimizer->SetRelaxationFactor( 0.7 );
optimizer->SetNumberOfIterations( 50 );
// Software Guide : EndCodeSnippet
// Optionally, get the step length from the command line arguments
if( argc > 11 )
{
optimizer->SetMaximumStepLength( std::stod( argv[12] ) );
}
// Optionally, get the number of iterations from the command line arguments
if( argc > 12 )
{
optimizer->SetNumberOfIterations( std::stoi( argv[13] ) );
}
//
// The BSpline transform has a large number of parameters, we use therefore a
// much larger number of samples to run this stage.
//
// Regulating the number of samples in the Metric is equivalent to performing
// multi-resolution registration because it is indeed a sub-sampling of the
// image.
metric->SetNumberOfSpatialSamples( numberOfBSplineParameters * 100 );
std::cout << std::endl << "Starting Deformable Registration Coarse Grid" << std::endl;
try
{
memorymeter.Start( "Deformable Registration Coarse" );
chronometer.Start( "Deformable Registration Coarse" );
registration->Update();
chronometer.Stop( "Deformable Registration Coarse" );
memorymeter.Stop( "Deformable Registration Coarse" );
}
catch( itk::ExceptionObject & err )
{
std::cerr << "ExceptionObject caught !" << std::endl;
std::cerr << err << std::endl;
return EXIT_FAILURE;
}
std::cout << "Deformable Registration Coarse Grid completed" << std::endl;
std::cout << std::endl;
OptimizerType::ParametersType finalParameters =
registration->GetLastTransformParameters();
bsplineTransformCoarse->SetParameters( finalParameters );
// Software Guide : BeginLatex
//
// Once the registration has finished with the low resolution grid, we
// proceed to instantiate a higher resolution
// \code{BSplineTransform}.
//
// Software Guide : EndLatex
DeformableTransformType::Pointer bsplineTransformFine = DeformableTransformType::New();
unsigned int numberOfGridNodesInOneDimensionFine = 5;
meshSize.Fill( numberOfGridNodesInOneDimensionFine - SplineOrder );
bsplineTransformFine->SetTransformDomainOrigin( fixedOrigin );
bsplineTransformFine->SetTransformDomainPhysicalDimensions(
fixedPhysicalDimensions );
bsplineTransformFine->SetTransformDomainMeshSize( meshSize );
bsplineTransformFine->SetTransformDomainDirection(
fixedImage->GetDirection() );
numberOfBSplineParameters = bsplineTransformFine->GetNumberOfParameters();
ParametersType parametersHigh( numberOfBSplineParameters );
parametersHigh.Fill( 0.0 );
// Software Guide : BeginLatex
//
// Now we need to initialize the BSpline coefficients of the higher resolution
// transform. This is done by first computing the actual deformation field
// at the higher resolution from the lower resolution BSpline coefficients.
// Then a BSpline decomposition is done to obtain the BSpline coefficient of
// the higher resolution transform.
//
// Software Guide : EndLatex
unsigned int counter = 0;
for ( unsigned int k = 0; k < SpaceDimension; k++ )
{
using ParametersImageType = DeformableTransformType::ImageType;
ResamplerType::Pointer upsampler = ResamplerType::New();
FunctionType::Pointer function = FunctionType::New();
upsampler->SetInput( bsplineTransformCoarse->GetCoefficientImages()[k] );
upsampler->SetInterpolator( function );
upsampler->SetTransform( identityTransform );
upsampler->SetSize( bsplineTransformFine->GetCoefficientImages()[k]->
GetLargestPossibleRegion().GetSize() );
upsampler->SetOutputSpacing( bsplineTransformFine->GetCoefficientImages()[k]->
GetSpacing() );
upsampler->SetOutputOrigin( bsplineTransformFine->GetCoefficientImages()[k]->
GetOrigin() );
using DecompositionType =
DecompositionType::Pointer decomposition = DecompositionType::New();
decomposition->SetSplineOrder( SplineOrder );
decomposition->SetInput( upsampler->GetOutput() );
decomposition->Update();
ParametersImageType::Pointer newCoefficients = decomposition->GetOutput();
// copy the coefficients into the parameter array
Iterator it( newCoefficients, bsplineTransformFine->GetCoefficientImages()[k]->
GetLargestPossibleRegion() );
while ( !it.IsAtEnd() )
{
parametersHigh[ counter++ ] = it.Get();
++it;
}
}
optimizerScales = OptimizerScalesType( numberOfBSplineParameters );
optimizerScales.Fill( 1.0 );
optimizer->SetScales( optimizerScales );
bsplineTransformFine->SetParameters( parametersHigh );
// Software Guide : BeginLatex
//
// We now pass the parameters of the high resolution transform as the initial
// parameters to be used in a second stage of the registration process.
//
// Software Guide : EndLatex
std::cout << "Starting Registration with high resolution transform" << std::endl;
// Software Guide : BeginCodeSnippet
registration->SetInitialTransformParameters(
bsplineTransformFine->GetParameters() );
registration->SetTransform( bsplineTransformFine );
//
// The BSpline transform at fine scale has a very large number of parameters,
// we use therefore a much larger number of samples to run this stage. In
// this case, however, the number of transform parameters is closer to the
// number of pixels in the image. Therefore we use the geometric mean of the
// two numbers to ensure that the number of samples is larger than the number
// of transform parameters and smaller than the number of samples.
//
// Regulating the number of samples in the Metric is equivalent to performing
// multi-resolution registration because it is indeed a sub-sampling of the
// image.
const unsigned long numberOfSamples =
static_cast<unsigned long>(
std::sqrt( static_cast<double>( numberOfBSplineParameters ) *
static_cast<double>( numberOfPixels ) ) );
metric->SetNumberOfSpatialSamples( numberOfSamples );
try
{
memorymeter.Start( "Deformable Registration Fine" );
chronometer.Start( "Deformable Registration Fine" );
registration->Update();
chronometer.Stop( "Deformable Registration Fine" );
memorymeter.Stop( "Deformable Registration Fine" );
}
catch( itk::ExceptionObject & err )
{
std::cerr << "ExceptionObject caught !" << std::endl;
std::cerr << err << std::endl;
return EXIT_FAILURE;
}
// Software Guide : EndCodeSnippet
std::cout << "Deformable Registration Fine Grid completed" << std::endl;
std::cout << std::endl;
// Report the time and memory taken by the registration
chronometer.Report( std::cout );
memorymeter.Report( std::cout );
finalParameters = registration->GetLastTransformParameters();
bsplineTransformFine->SetParameters( finalParameters );
using ResampleFilterType = itk::ResampleImageFilter<
MovingImageType,
FixedImageType >;
ResampleFilterType::Pointer resample = ResampleFilterType::New();
resample->SetTransform( bsplineTransformFine );
resample->SetInput( movingImageReader->GetOutput() );
resample->SetSize( fixedImage->GetLargestPossibleRegion().GetSize() );
resample->SetOutputOrigin( fixedImage->GetOrigin() );
resample->SetOutputSpacing( fixedImage->GetSpacing() );
resample->SetOutputDirection( fixedImage->GetDirection() );
// This value is set to zero in order to make easier to perform
// regression testing in this example. However, for didactic
// exercise it will be better to set it to a medium gray value
// such as 100 or 128.
resample->SetDefaultPixelValue( 0 );
using OutputPixelType = signed short;
using CastFilterType = itk::CastImageFilter<
FixedImageType,
OutputImageType >;
WriterType::Pointer writer = WriterType::New();
CastFilterType::Pointer caster = CastFilterType::New();
writer->SetFileName( argv[3] );
caster->SetInput( resample->GetOutput() );
writer->SetInput( caster->GetOutput() );
std::cout << "Writing resampled moving image...";
try
{
writer->Update();
}
catch( itk::ExceptionObject & err )
{
std::cerr << "ExceptionObject caught !" << std::endl;
std::cerr << err << std::endl;
return EXIT_FAILURE;
}
std::cout << " Done!" << std::endl;
using DifferenceFilterType = itk::SquaredDifferenceImageFilter<
FixedImageType,
FixedImageType,
OutputImageType >;
DifferenceFilterType::Pointer difference = DifferenceFilterType::New();
WriterType::Pointer writer2 = WriterType::New();
writer2->SetInput( difference->GetOutput() );
// Compute the difference image between the
// fixed and resampled moving image.
if( argc > 4 )
{
difference->SetInput1( fixedImageReader->GetOutput() );
difference->SetInput2( resample->GetOutput() );
writer2->SetFileName( argv[4] );
std::cout << "Writing difference image after registration...";
try
{
writer2->Update();
}
catch( itk::ExceptionObject & err )
{
std::cerr << "ExceptionObject caught !" << std::endl;
std::cerr << err << std::endl;
return EXIT_FAILURE;
}
std::cout << " Done!" << std::endl;
}
// Compute the difference image between the
// fixed and moving image before registration.
if( argc > 5 )
{
writer2->SetFileName( argv[5] );
difference->SetInput1( fixedImageReader->GetOutput() );
resample->SetTransform( identityTransform );
std::cout << "Writing difference image before registration...";
try
{
writer2->Update();
}
catch( itk::ExceptionObject & err )
{
std::cerr << "ExceptionObject caught !" << std::endl;
std::cerr << err << std::endl;
return EXIT_FAILURE;
}
std::cout << " Done!" << std::endl;
}
// Generate the explicit deformation field resulting from
// the registration.
if( argc > 6 )
{
using DisplacementFieldType = itk::Image< VectorType, ImageDimension >;
DisplacementFieldType::Pointer field = DisplacementFieldType::New();
field->SetRegions( fixedRegion );
field->SetOrigin( fixedImage->GetOrigin() );
field->SetSpacing( fixedImage->GetSpacing() );
field->SetDirection( fixedImage->GetDirection() );
field->Allocate();
FieldIterator fi( field, fixedRegion );
fi.GoToBegin();
DeformableTransformType::InputPointType fixedPoint;
DeformableTransformType::OutputPointType movingPoint;
VectorType displacement;
while( ! fi.IsAtEnd() )
{
index = fi.GetIndex();
field->TransformIndexToPhysicalPoint( index, fixedPoint );
movingPoint = bsplineTransformFine->TransformPoint( fixedPoint );
displacement = movingPoint - fixedPoint;
fi.Set( displacement );
++fi;
}
FieldWriterType::Pointer fieldWriter = FieldWriterType::New();
fieldWriter->SetInput( field );
fieldWriter->SetFileName( argv[6] );
std::cout << "Writing deformation field ...";
try
{
fieldWriter->Update();
}
catch( itk::ExceptionObject & excp )
{
std::cerr << "Exception thrown " << std::endl;
std::cerr << excp << std::endl;
return EXIT_FAILURE;
}
std::cout << " Done!" << std::endl;
}
// Optionally, save the transform parameters in a file
if( argc > 9 )
{
std::cout << "Writing transform parameter file ...";
std::ofstream parametersFile;
parametersFile.open( argv[9] );
parametersFile << finalParameters << std::endl;
parametersFile.close();
std::cout << " Done!" << std::endl;
}
return EXIT_SUCCESS;
}