C API Reference¶

Functions declared in parallelproj.h.

Functions

int parallelproj_cuda_enabled(void)¶

Query whether this libparallelproj library was compiled with CUDA support.

Returns:: 1 if compiled with CUDA support, 0 otherwise.

const char *parallelproj_version(void)¶

Query the version string of this libparallelproj library. Potentially includes extra suffix for “dirty” versions.

Returns:: Null-terminated version string.

int parallelproj_version_major(void)¶

Query the major version number of this libparallelproj library.

Returns:: Major version number.

int parallelproj_version_minor(void)¶

Query the minor version number of this libparallelproj library.

Returns:: Minor version number.

int parallelproj_version_patch(void)¶

Query the patch version number of this libparallelproj library.

Returns:: Patch version number.

void joseph3d_fwd(const float *lor_start, const float *lor_end, const float *image, const float *image_origin, const float *voxel_size, float *projection_values, std::size_t num_lors, const int *image_dim, int device_id = 0, int threads_per_block = 64)¶

Forward projection using the Joseph 3D algorithm.

Note

All pointers can be host pointers, CUDA device pointers, or CUDA managed pointers.

Note

Image memory layout for image is C-contiguous with index n1*n2*i + n2*j + k for voxel/pixel [i,j,k].

Parameters:

lor_start – Pointer to array of shape [3 * num_lors] with the coordinates of the start points of the LORs/events. The start coordinates of the n-th LOR/event are at lor_start[n*3 + i] with i = 0,1,2. Units are those of voxel_size.
lor_end – Pointer to array of shape [3 * num_lors] with the coordinates of the end points of the LORs/events. The end coordinates of the n-th LOR/event are at lor_end[n*3 + i] with i = 0,1,2. Units are those of voxel_size.
image – Pointer to array of shape [n0*n1*n2] containing the 3D image used for forward projection.
image_origin – Pointer to array [x0_0, x0_1, x0_2] of coordinates of the center of the [0,0,0] voxel.
voxel_size – Pointer to array [vs0, vs1, vs2] of voxel sizes.
projection_values – Pointer to array of length num_lors where the forward projection results will be stored.
num_lors – Number of geometric LORs.
image_dim – Pointer to array with image dimensions [n0, n1, n2].
device_id – ID of the device to use for computation (default: 0).
threads_per_block – Number of threads per block for GPU computation (default: 64).

void joseph3d_back(const float *lor_start, const float *lor_end, float *image, const float *image_origin, const float *voxel_size, const float *projection_values, std::size_t num_lors, const int *image_dim, int device_id = 0, int threads_per_block = 64)¶

Backprojection using the Joseph 3D algorithm.

Note

All pointers can be host pointers, CUDA device pointers, or CUDA managed pointers.

Note

Image memory layout for image is C-contiguous with index n1*n2*i + n2*j + k for voxel/pixel [i,j,k].

Parameters:

lor_start – Pointer to array of shape [3 * num_lors] with the coordinates of the start points of the LORs/events. The start coordinates of the n-th LOR/event are at lor_start[n*3 + i] with i = 0,1,2. Units are those of voxel_size.
lor_end – Pointer to array of shape [3 * num_lors] with the coordinates of the end points of the LORs/events. The end coordinates of the n-th LOR/event are at lor_end[n*3 + i] with i = 0,1,2. Units are those of voxel_size.
image – Pointer to array of shape [n0*n1*n2] containing the 3D image used for backprojection. Values are added to the existing contents of this array.
image_origin – Pointer to array [x0_0, x0_1, x0_2] of coordinates of the center of the [0,0,0] voxel.
voxel_size – Pointer to array [vs0, vs1, vs2] of voxel sizes.
projection_values – Pointer to array of length num_lors with the values to be backprojected.
num_lors – Number of geometric LORs.
image_dim – Pointer to array with image dimensions [n0, n1, n2].
device_id – ID of the device to use for computation (default: 0).
threads_per_block – Number of threads per block for GPU computation (default: 64).

void joseph3d_tof_sino_fwd(const float *lor_start, const float *lor_end, const float *image, const float *image_origin, const float *voxel_size, float *projection_values, std::size_t num_lors, const int *image_dim, float tof_bin_width, const float *tof_sigma, const float *tof_center_offset, float num_sigmas, short num_tof_bins, unsigned char is_lor_dependent_tof_sigma, unsigned char is_lor_dependent_tof_center_offset, int device_id = 0, int threads_per_block = 64)¶

3D sinogram TOF Joseph forward projector

Note

All pointers can be host pointers, CUDA device pointers, or CUDA managed pointers.

Note

Image memory layout for image is C-contiguous with index n1*n2*i + n2*j + k for voxel/pixel [i,j,k].

Note

TOF sinogram memory layout for projection_values is row-major per LOR: [LOR0-TOFBIN-0, LOR0-TOFBIN-1, …, LOR0-TOFBIN-(num_tof_bins-1), LOR1-TOFBIN-0, LOR1-TOFBIN-1, …, LOR1-TOFBIN-(num_tof_bins-1), …].

Parameters:

lor_start – Pointer to array of shape [3 * num_lors] with the coordinates of the start points of the LORs/events. The start coordinates of the n-th LOR/event are at lor_start[n*3 + i] with i = 0,1,2. Units are those of voxel_size.
lor_end – Pointer to array of shape [3 * num_lors] with the coordinates of the end points of the LORs/events. The end coordinates of the n-th LOR/event are at lor_end[n*3 + i] with i = 0,1,2. Units are those of voxel_size.
image – Pointer to array of shape [n0*n1*n2] containing the 3D image used for forward projection.
image_origin – Pointer to array [x0_0, x0_1, x0_2] of coordinates of the center of the [0,0,0] voxel.
voxel_size – Pointer to array [vs0, vs1, vs2] of voxel sizes.
projection_values – Pointer to array of length num_lors * num_tof_bins (output) used to store the projections.
num_lors – Number of geometric LORs.
image_dim – Pointer to array with image dimensions [n0, n1, n2].
tof_bin_width – Width of the TOF bins in spatial units (units of lor_start and lor_end).
tof_sigma – Pointer to array of length 1 or num_lors (depending on is_lor_dependent_tof_sigma). Contains TOF resolution (sigma) values in spatial units (units of lor_start and lor_end).
tof_center_offset – Pointer to array of length 1 or num_lors (depending on is_lor_dependent_tof_center_offset). Contains offsets of the central TOF bin from the LOR midpoint in spatial units. A positive value means a shift towards the end point of the LOR.
num_sigmas – Number of sigmas to consider for calculation of the TOF kernel.
num_tof_bins – Number of TOF bins.
is_lor_dependent_tof_sigma – Unsigned char 0 or 1. 0 means that the first value in tof_sigma is used for all LORs. 1 (non-zero) means that the TOF resolutions are LOR dependent.
is_lor_dependent_tof_center_offset – Unsigned char 0 or 1. 0 means that the first value in tof_center_offset is used for all LORs. 1 (non-zero) means that the TOF center offsets are LOR dependent.
device_id – ID of the device to use for computation (default: 0).
threads_per_block – Number of threads per block for GPU computation (default: 64).

void joseph3d_tof_sino_back(const float *lor_start, const float *lor_end, float *image, const float *image_origin, const float *voxel_size, const float *projection_values, std::size_t num_lors, const int *image_dim, float tof_bin_width, const float *tof_sigma, const float *tof_center_offset, float num_sigmas, short num_tof_bins, unsigned char is_lor_dependent_tof_sigma, unsigned char is_lor_dependent_tof_center_offset, int device_id = 0, int threads_per_block = 64)¶

TOF sinogram backprojection using the Joseph 3D algorithm.

The function backprojects a TOF sinogram into a 3D image volume using the Joseph ray-driven algorithm. Values from projection_values are distributed into image (accumulated, not overwritten).

Note

All pointers can be host pointers, CUDA device pointers, or CUDA managed pointers.

Note

Image memory layout for image is C-contiguous with index n1*n2*i + n2*j + k for voxel/pixel [i,j,k].

Note

Parameters:

lor_start – Pointer to array of shape [3 * num_lors] with the coordinates of the start points of the LORs/events. The start coordinates of the n-th LOR/event are at lor_start[n*3 + i] with i = 0,1,2. Units are those of voxel_size.
lor_end – Pointer to array of shape [3 * num_lors] with the coordinates of the end points of the LORs/events. The end coordinates of the n-th LOR/event are at lor_end[n*3 + i] with i = 0,1,2. Units are those of voxel_size.
image – Pointer to array of shape [n0*n1*n2] containing the 3D image used for backprojection. Values are added to the existing contents of this array.
image_origin – Pointer to array [x0_0, x0_1, x0_2] of coordinates of the center of the [0,0,0] voxel.
voxel_size – Pointer to array [vs0, vs1, vs2] of voxel sizes.
projection_values – Pointer to TOF sinogram data of length num_lors * num_tof_bins (see details).
num_lors – Number of geometric LORs.
image_dim – Pointer to array with image dimensions [n0, n1, n2].
tof_bin_width – Width of the TOF bins in spatial units (units of lor_start and lor_end).
tof_sigma – Pointer to array of length 1 or num_lors (depending on is_lor_dependent_tof_sigma). Contains TOF resolution (sigma) values in spatial units (units of lor_start and lor_end).
tof_center_offset – Pointer to array of length 1 or num_lors (depending on is_lor_dependent_tof_center_offset). Contains offsets of the central TOF bin from the LOR midpoint in spatial units. A positive value means a shift towards the end point of the LOR.
num_sigmas – Number of sigmas to consider for calculation of the TOF kernel.
num_tof_bins – Number of TOF bins.
is_lor_dependent_tof_sigma – Unsigned char 0 or 1. 0 means that the first value in tof_sigma is used for all LORs. 1 (non-zero) means that the TOF resolutions are LOR dependent.
is_lor_dependent_tof_center_offset – Unsigned char 0 or 1. 0 means that the first value in tof_center_offset is used for all LORs. 1 (non-zero) means that the TOF center offsets are LOR dependent.
device_id – ID of the device to use for computation (default: 0).
threads_per_block – Number of threads per block for GPU computation (default: 64).

void joseph3d_tof_lm_fwd(const float *lor_start, const float *lor_end, const float *image, const float *image_origin, const float *voxel_size, float *projection_values, std::size_t num_events, const int *image_dim, float tof_bin_width, const float *tof_sigma, const float *tof_center_offset, float num_sigmas, const short *tof_bin_index, short num_tof_bins, unsigned char is_lor_dependent_tof_sigma, unsigned char is_lor_dependent_tof_center_offset, int device_id = 0, int threads_per_block = 64)¶

3D listmode TOF Joseph forward projector

Note

All pointers can be host pointers, CUDA device pointers, or CUDA managed pointers.

Note

Image memory layout for image is C-contiguous with index n1*n2*i + n2*j + k for voxel/pixel [i,j,k].

Parameters:

lor_start – Pointer to array of shape [3 * num_events] with the coordinates of the start points of the LORs/events. The start coordinates of the n-th LOR/event are at lor_start[n*3 + i] with i = 0,1,2. Units are those of voxel_size.
lor_end – Pointer to array of shape [3 * num_events] with the coordinates of the end points of the LORs/events. The end coordinates of the n-th LOR/event are at lor_end[n*3 + i] with i = 0,1,2. Units are those of voxel_size.
image – Pointer to array of shape [n0*n1*n2] containing the 3D image used for forward projection.
image_origin – Pointer to array [x0_0, x0_1, x0_2] of coordinates of the center of the [0,0,0] voxel.
voxel_size – Pointer to array [vs0, vs1, vs2] of voxel sizes.
projection_values – Pointer to array of length num_events (output) used to store the projections.
num_events – Number of events.
image_dim – Pointer to array with image dimensions [n0, n1, n2].
tof_bin_width – Width of the TOF bins in spatial units (units of lor_start and lor_end).
tof_sigma – Pointer to array of length 1 or num_events (depending on is_lor_dependent_tof_sigma). Contains TOF resolution (sigma) values in spatial units (units of lor_start and lor_end).
tof_center_offset – Pointer to array of length 1 or num_events (depending on is_lor_dependent_tof_center_offset). Contains offsets of the central TOF bin from the LOR midpoint in spatial units. A positive value means a shift towards the end point of the LOR.
num_sigmas – Number of sigmas to consider for calculation of the TOF kernel.
tof_bin_index – Pointer to array of length num_events with the TOF bin numbers.
num_tof_bins – Number of TOF bins.
is_lor_dependent_tof_sigma – Unsigned char 0 or 1. 0 means that the first value in tof_sigma is used for all events. 1 (non-zero) means that the TOF resolutions are LOR dependent.
is_lor_dependent_tof_center_offset – Unsigned char 0 or 1. 0 means that the first value in tof_center_offset is used for all events. 1 (non-zero) means that the TOF center offsets are LOR dependent.
device_id – ID of the device to use for computation (default: 0).
threads_per_block – Number of threads per block for GPU computation (default: 64).

void joseph3d_tof_lm_back(const float *lor_start, const float *lor_end, float *image, const float *image_origin, const float *voxel_size, const float *projection_values, std::size_t num_events, const int *image_dim, float tof_bin_width, const float *tof_sigma, const float *tof_center_offset, float num_sigmas, const short *tof_bin_index, short num_tof_bins, unsigned char is_lor_dependent_tof_sigma, unsigned char is_lor_dependent_tof_center_offset, int device_id = 0, int threads_per_block = 64)¶

3D listmode TOF Joseph back projector

Note

All pointers can be host pointers, CUDA device pointers, or CUDA managed pointers.

Note

Image memory layout for image is C-contiguous with index n1*n2*i + n2*j + k for voxel/pixel [i,j,k].

Parameters:

lor_start – Pointer to array of shape [3 * num_events] with the coordinates of the start points of the LORs/events. The start coordinates of the n-th LOR/event are at lor_start[n*3 + i] with i = 0,1,2. Units are those of voxel_size.
lor_end – Pointer to array of shape [3 * num_events] with the coordinates of the end points of the LORs/events. The end coordinates of the n-th LOR/event are at lor_end[n*3 + i] with i = 0,1,2. Units are those of voxel_size.
image – Pointer to array of shape [n0*n1*n2] containing the 3D image used for backprojection. Values are added to the existing contents of this array.
image_origin – Pointer to array [x0_0, x0_1, x0_2] of coordinates of the center of the [0,0,0] voxel.
voxel_size – Pointer to array [vs0, vs1, vs2] of voxel sizes.
projection_values – Pointer to array of values to be backprojected (length num_events).
num_events – Number of events.
image_dim – Pointer to array with image dimensions [n0, n1, n2].
tof_bin_width – Width of the TOF bins in spatial units (units of lor_start and lor_end).
tof_sigma – Pointer to array of length 1 or num_events (depending on is_lor_dependent_tof_sigma). Contains TOF resolution (sigma) values in spatial units (units of lor_start and lor_end).
tof_center_offset – Pointer to array of length 1 or num_events (depending on is_lor_dependent_tof_center_offset). Contains offsets of the central TOF bin from the LOR midpoint in spatial units. A positive value means a shift towards the end point of the LOR.
num_sigmas – Number of sigmas to consider for calculation of the TOF kernel.
tof_bin_index – Pointer to array of length num_events with the TOF bin numbers.
num_tof_bins – Number of TOF bins.
is_lor_dependent_tof_sigma – Unsigned char 0 or 1. 0 means that the first value in tof_sigma is used for all events. 1 (non-zero) means that the TOF resolutions are LOR dependent.
is_lor_dependent_tof_center_offset – Unsigned char 0 or 1. 0 means that the first value in tof_center_offset is used for all events. 1 (non-zero) means that the TOF center offsets are LOR dependent.
device_id – ID of the device to use for computation (default: 0).
threads_per_block – Number of threads per block for GPU computation (default: 64).