High throughput CUDA implementation of accurate geometric modelling for iterative reconstruction of PET data

P. J. Markiewicz, K. Thielemans, M. J. Ehrhardt, J. Jiao, N. Burgos, David Atkinson, S. R. Arridge, B. F. Hutton, S. Ourselin

Research output: Chapter in Book/Report/Conference proceedingConference contribution

4 Citations (Scopus)

Abstract

An approach to high throughput and high accuracy modelling of the geometric component of PET acquisition for the Siemens Biograph mMR PET/MR scanner is presented. The geometric components calculated in forward and back-projections are computationally expensive, however, they are inherently parallel and therefore, they are suitable for implementation on parallel computing platforms such as CUDA, consequently permitting more accurate and computationally involved system models. The key aspects of this work are: (1) accurate modelling of the geometric component of each tube of response (TOR) by tracing multiple lines for each TOR to account for the varying sensitivity along and across the TOR; (2) decomposition of the calculations into transaxial and axial components allowing the use of ray and voxel-driven methods optimal for forward and back-projection, respectively. Such decomposition also allows keeping exact correspondence between the ray and voxel-driven methods for forward and backprojection. (3) Due to the large axial field of view and the high number of crystal rings (64), the core ray-tracing is taking place in the axial dimension by projecting the transaxial calculations on each direct and oblique sinogram element. (4) Axially oriented arrangement of the images and sinograms in the GPU memory enables more efficient use of the L2 cache and together with point (3) it leads to more optimal performance even without the use of shared memory. Currently, for the Biograph mMR scanner geometry the projections are calculated within two seconds.

Original languageEnglish
Title of host publication2014 IEEE Nuclear Science Symposium and Medical Imaging Conference, NSS/MIC
Place of PublicationU. S. A.
PublisherIEEE
ISBN (Electronic)9781479960972
DOIs
Publication statusPublished - 14 Mar 2016
EventIEEE Nuclear Science Symposium and Medical Imaging Conference, NSS/MIC 2014 - Seattle, USA United States
Duration: 8 Nov 201415 Nov 2014

Conference

ConferenceIEEE Nuclear Science Symposium and Medical Imaging Conference, NSS/MIC 2014
CountryUSA United States
CitySeattle
Period8/11/1415/11/14

ASJC Scopus subject areas

  • Nuclear and High Energy Physics
  • Radiology Nuclear Medicine and imaging

Cite this

Markiewicz, P. J., Thielemans, K., Ehrhardt, M. J., Jiao, J., Burgos, N., Atkinson, D., ... Ourselin, S. (2016). High throughput CUDA implementation of accurate geometric modelling for iterative reconstruction of PET data. In 2014 IEEE Nuclear Science Symposium and Medical Imaging Conference, NSS/MIC [7430963] U. S. A.: IEEE. https://doi.org/10.1109/NSSMIC.2014.7430963

High throughput CUDA implementation of accurate geometric modelling for iterative reconstruction of PET data. / Markiewicz, P. J.; Thielemans, K.; Ehrhardt, M. J.; Jiao, J.; Burgos, N.; Atkinson, David; Arridge, S. R.; Hutton, B. F.; Ourselin, S.

2014 IEEE Nuclear Science Symposium and Medical Imaging Conference, NSS/MIC . U. S. A. : IEEE, 2016. 7430963.

Research output: Chapter in Book/Report/Conference proceedingConference contribution

Markiewicz, PJ, Thielemans, K, Ehrhardt, MJ, Jiao, J, Burgos, N, Atkinson, D, Arridge, SR, Hutton, BF & Ourselin, S 2016, High throughput CUDA implementation of accurate geometric modelling for iterative reconstruction of PET data. in 2014 IEEE Nuclear Science Symposium and Medical Imaging Conference, NSS/MIC ., 7430963, IEEE, U. S. A., IEEE Nuclear Science Symposium and Medical Imaging Conference, NSS/MIC 2014, Seattle, USA United States, 8/11/14. https://doi.org/10.1109/NSSMIC.2014.7430963
Markiewicz PJ, Thielemans K, Ehrhardt MJ, Jiao J, Burgos N, Atkinson D et al. High throughput CUDA implementation of accurate geometric modelling for iterative reconstruction of PET data. In 2014 IEEE Nuclear Science Symposium and Medical Imaging Conference, NSS/MIC . U. S. A.: IEEE. 2016. 7430963 https://doi.org/10.1109/NSSMIC.2014.7430963
Markiewicz, P. J. ; Thielemans, K. ; Ehrhardt, M. J. ; Jiao, J. ; Burgos, N. ; Atkinson, David ; Arridge, S. R. ; Hutton, B. F. ; Ourselin, S. / High throughput CUDA implementation of accurate geometric modelling for iterative reconstruction of PET data. 2014 IEEE Nuclear Science Symposium and Medical Imaging Conference, NSS/MIC . U. S. A. : IEEE, 2016.
@inproceedings{21837a540cbb4541abf0e62a124be3ff,
title = "High throughput CUDA implementation of accurate geometric modelling for iterative reconstruction of PET data",
abstract = "An approach to high throughput and high accuracy modelling of the geometric component of PET acquisition for the Siemens Biograph mMR PET/MR scanner is presented. The geometric components calculated in forward and back-projections are computationally expensive, however, they are inherently parallel and therefore, they are suitable for implementation on parallel computing platforms such as CUDA, consequently permitting more accurate and computationally involved system models. The key aspects of this work are: (1) accurate modelling of the geometric component of each tube of response (TOR) by tracing multiple lines for each TOR to account for the varying sensitivity along and across the TOR; (2) decomposition of the calculations into transaxial and axial components allowing the use of ray and voxel-driven methods optimal for forward and back-projection, respectively. Such decomposition also allows keeping exact correspondence between the ray and voxel-driven methods for forward and backprojection. (3) Due to the large axial field of view and the high number of crystal rings (64), the core ray-tracing is taking place in the axial dimension by projecting the transaxial calculations on each direct and oblique sinogram element. (4) Axially oriented arrangement of the images and sinograms in the GPU memory enables more efficient use of the L2 cache and together with point (3) it leads to more optimal performance even without the use of shared memory. Currently, for the Biograph mMR scanner geometry the projections are calculated within two seconds.",
author = "Markiewicz, {P. J.} and K. Thielemans and Ehrhardt, {M. J.} and J. Jiao and N. Burgos and David Atkinson and Arridge, {S. R.} and Hutton, {B. F.} and S. Ourselin",
year = "2016",
month = "3",
day = "14",
doi = "10.1109/NSSMIC.2014.7430963",
language = "English",
booktitle = "2014 IEEE Nuclear Science Symposium and Medical Imaging Conference, NSS/MIC",
publisher = "IEEE",
address = "USA United States",

}

TY - GEN

T1 - High throughput CUDA implementation of accurate geometric modelling for iterative reconstruction of PET data

AU - Markiewicz, P. J.

AU - Thielemans, K.

AU - Ehrhardt, M. J.

AU - Jiao, J.

AU - Burgos, N.

AU - Atkinson, David

AU - Arridge, S. R.

AU - Hutton, B. F.

AU - Ourselin, S.

PY - 2016/3/14

Y1 - 2016/3/14

N2 - An approach to high throughput and high accuracy modelling of the geometric component of PET acquisition for the Siemens Biograph mMR PET/MR scanner is presented. The geometric components calculated in forward and back-projections are computationally expensive, however, they are inherently parallel and therefore, they are suitable for implementation on parallel computing platforms such as CUDA, consequently permitting more accurate and computationally involved system models. The key aspects of this work are: (1) accurate modelling of the geometric component of each tube of response (TOR) by tracing multiple lines for each TOR to account for the varying sensitivity along and across the TOR; (2) decomposition of the calculations into transaxial and axial components allowing the use of ray and voxel-driven methods optimal for forward and back-projection, respectively. Such decomposition also allows keeping exact correspondence between the ray and voxel-driven methods for forward and backprojection. (3) Due to the large axial field of view and the high number of crystal rings (64), the core ray-tracing is taking place in the axial dimension by projecting the transaxial calculations on each direct and oblique sinogram element. (4) Axially oriented arrangement of the images and sinograms in the GPU memory enables more efficient use of the L2 cache and together with point (3) it leads to more optimal performance even without the use of shared memory. Currently, for the Biograph mMR scanner geometry the projections are calculated within two seconds.

AB - An approach to high throughput and high accuracy modelling of the geometric component of PET acquisition for the Siemens Biograph mMR PET/MR scanner is presented. The geometric components calculated in forward and back-projections are computationally expensive, however, they are inherently parallel and therefore, they are suitable for implementation on parallel computing platforms such as CUDA, consequently permitting more accurate and computationally involved system models. The key aspects of this work are: (1) accurate modelling of the geometric component of each tube of response (TOR) by tracing multiple lines for each TOR to account for the varying sensitivity along and across the TOR; (2) decomposition of the calculations into transaxial and axial components allowing the use of ray and voxel-driven methods optimal for forward and back-projection, respectively. Such decomposition also allows keeping exact correspondence between the ray and voxel-driven methods for forward and backprojection. (3) Due to the large axial field of view and the high number of crystal rings (64), the core ray-tracing is taking place in the axial dimension by projecting the transaxial calculations on each direct and oblique sinogram element. (4) Axially oriented arrangement of the images and sinograms in the GPU memory enables more efficient use of the L2 cache and together with point (3) it leads to more optimal performance even without the use of shared memory. Currently, for the Biograph mMR scanner geometry the projections are calculated within two seconds.

UR - http://www.scopus.com/inward/record.url?scp=84965025863&partnerID=8YFLogxK

U2 - 10.1109/NSSMIC.2014.7430963

DO - 10.1109/NSSMIC.2014.7430963

M3 - Conference contribution

BT - 2014 IEEE Nuclear Science Symposium and Medical Imaging Conference, NSS/MIC

PB - IEEE

CY - U. S. A.

ER -