#include <rtkProjectionStackToFourDImageFilter.h>

Inheritance diagram for rtk::ProjectionStackToFourDImageFilter< VolumeSeriesType, ProjectionStackType, TFFTPrecision >:

Collaboration diagram for rtk::ProjectionStackToFourDImageFilter< VolumeSeriesType, ProjectionStackType, TFFTPrecision >:

Public Types
using	BackProjectionFilterType = rtk::BackProjectionImageFilter< VolumeType, VolumeType >

using	ConstantVolumeSeriesSourceType = rtk::ConstantImageSource< VolumeSeriesType >

using	ConstantVolumeSourceType = rtk::ConstantImageSource< VolumeType >

using	CPUVolumeSeriesType = typename itk::Image< typename VolumeSeriesType::PixelType, VolumeSeriesType::ImageDimension >

using	CudaConstantVolumeSeriesSourceType = typename std::conditional_t< std::is_same_v< VolumeSeriesType, CPUVolumeSeriesType >, ConstantVolumeSeriesSourceType, CudaConstantVolumeSeriesSource >

using	CudaConstantVolumeSourceType = typename std::conditional_t< std::is_same_v< VolumeSeriesType, CPUVolumeSeriesType >, ConstantVolumeSourceType, CudaConstantVolumeSource >

using	CudaSplatImageFilterType = typename std::conditional_t< std::is_same_v< VolumeSeriesType, CPUVolumeSeriesType >, SplatFilterType, CudaSplatImageFilter >

using	ExtractFilterType = itk::ExtractImageFilter< ProjectionStackType, ProjectionStackType >

using	GeometryType = rtk::ThreeDCircularProjectionGeometry

using	Pointer = itk::SmartPointer< Self >

using	Self = ProjectionStackToFourDImageFilter

using	SplatFilterType = rtk::SplatWithKnownWeightsImageFilter< VolumeSeriesType, VolumeType >

using	Superclass = itk::ImageToImageFilter< VolumeSeriesType, VolumeSeriesType >

using	VolumeType = ProjectionStackType

Public Member Functions
virtual ::itk::LightObject::Pointer	CreateAnother () const

const char *	GetNameOfClass () const override

void	SetBackProjectionFilter (BackProjectionFilterType *_arg)

virtual void	SetGeometry (const GeometryType *_arg)

virtual void	SetSignal (const std::vector< double > signal)


void	SetInputVolumeSeries (const VolumeSeriesType *VolumeSeries)

VolumeSeriesType::ConstPointer	GetInputVolumeSeries ()


void	SetInputProjectionStack (const ProjectionStackType *Projection)

ProjectionStackType::ConstPointer	GetInputProjectionStack ()


virtual void	SetUseCudaSplat (bool _arg)

virtual bool	GetUseCudaSplat ()

virtual void	SetUseCudaSources (bool _arg)

virtual bool	GetUseCudaSources ()


virtual itk::Array2D< float >	GetWeights ()

virtual void	SetWeights (itk::Array2D< float > _arg)

Static Public Member Functions
static Pointer	New ()

Protected Member Functions
void	GenerateData () override

void	GenerateInputRequestedRegion () override

void	GenerateOutputInformation () override

void	InitializeConstantSource ()

	ProjectionStackToFourDImageFilter ()

void	VerifyPreconditions () const override

	~ProjectionStackToFourDImageFilter () override=default

Protected Attributes
BackProjectionFilterType::Pointer	m_BackProjectionFilter

ConstantVolumeSeriesSourceType::Pointer	m_ConstantVolumeSeriesSource

ConstantVolumeSourceType::Pointer	m_ConstantVolumeSource

ExtractFilterType::Pointer	m_ExtractFilter

GeometryType::ConstPointer	m_Geometry

std::vector< double >	m_Signal

SplatFilterType::Pointer	m_SplatFilter

bool	m_UseCudaSources

bool	m_UseCudaSplat

itk::Array2D< float >	m_Weights

Detailed Description

template<typename VolumeSeriesType, typename ProjectionStackType, typename TFFTPrecision = double>
class rtk::ProjectionStackToFourDImageFilter< VolumeSeriesType, ProjectionStackType, TFFTPrecision >

Implements part of the 4D reconstruction by conjugate gradient.

See the reference paper: "Cardiac C-arm computed tomography using a 3D + time ROI reconstruction method with spatial and temporal regularization" by Mory et al.

4D conjugate gradient reconstruction consists in iteratively minimizing the following cost function:

Sum_over_theta || R_theta S_theta f - p_theta ||_2^2

with

f a 4D series of 3D volumes, each one being the reconstruction at a given respiratory/cardiac phase
p_theta is the projection measured at angle theta
S_theta an interpolation operator which, from the 3D + time sequence f, estimates the 3D volume through which projection p_theta has been acquired
R_theta is the X-ray transform (the forward projection operator) for angle theta

Computing the gradient of this cost function yields:

S_theta^T R_theta^T R_theta S_theta f - S_theta^T R_theta^T p_theta

where A^T means the adjoint of operator A.

ProjectionStackToFourDImageFilter implements S_theta^T R_theta^T.

Test:: rtkfourdconjugategradienttest.cxx

Author: Cyril Mory

Definition at line 105 of file rtkProjectionStackToFourDImageFilter.h.