Abstract
Computational acceleration on graphics processing units (GPUs) can make advanced magnetic resonance imaging (MRI) reconstruction algorithms attractive in clinical settings, thereby improving the quality of MR images across a broad spectrum of applications. At present, MR imaging is often limited by high noise levels, significant imaging artifacts, and/or long data acquisition (scan) times. Advanced image reconstruction algorithms can mitigate these limitations and improve image quality by simultaneously operating on scan data acquired with arbitrary trajectories and incorporating additional information such as anatomical constraints. However, the improvements in image quality come at the expense of a considerable increase in computation. This paper describes the acceleration of an advanced reconstruction algorithm on NVIDIA's Quadro FX 5600. Optimizations such as register allocating the voxel data, tiling the scan data, and storing the scan data in the Quadro's constant memory dramatically reduce the reconstruction's required bandwidth to off-chip memory. The Quadro's special functional units provide substantial acceleration of the trigonometric computations in the algorithm's inner loops, and experimentally-tuned code transformations increase the reconstruction's performance by an additional 20%. The reconstruction of a 3D image with 1283 voxels ultimately achieves 150 GFLOPS and requires less than two minutes on the Quadro, while reconstruction on a quadcore CPU is thirteen times slower. Furthermore, relative to the true image, the error exhibited by the advanced reconstruction is only 12%, while conventional reconstruction techniques incur error of 42%. In short, the acceleration afforded by the GPU greatly increases the appeal of the advanced reconstruction for clinical MRI applications.
| Original language | English (US) |
|---|---|
| Title of host publication | Conference on Computing Frontiers - Proceedings of the 2008 Conference on Computing Frontiers, CF'08 |
| Publisher | Association for Computing Machinery |
| Pages | 261-272 |
| Number of pages | 12 |
| ISBN (Print) | 9781605580777 |
| DOIs | |
| State | Published - Jan 1 2008 |
| Event | 2008 Conference on Computing Frontiers, CF'08 - Ischia, Italy Duration: May 5 2008 → May 7 2008 |
Publication series
| Name | Conference on Computing Frontiers - Proceedings of the 2008 Conference on Computing Frontiers, CF'08 |
|---|
Other
| Other | 2008 Conference on Computing Frontiers, CF'08 |
|---|---|
| Country | Italy |
| City | Ischia |
| Period | 5/5/08 → 5/7/08 |
Fingerprint
Keywords
- CUDA
- GPGPU
- GPU computing
- MRI
- Reconstruction
ASJC Scopus subject areas
- Computer Science Applications
- Hardware and Architecture
- Software
- Electrical and Electronic Engineering
Cite this
Accelerating advanced MRI reconstructions on GPUs. / Stone, Sam S.; Haldar, Justin P.; Tsao, Stephanie C.; Hwu, Wen Mei W; Liang, Zhi Pei; Sutton, Bradley P.
Conference on Computing Frontiers - Proceedings of the 2008 Conference on Computing Frontiers, CF'08. Association for Computing Machinery, 2008. p. 261-272 (Conference on Computing Frontiers - Proceedings of the 2008 Conference on Computing Frontiers, CF'08).Research output: Chapter in Book/Report/Conference proceeding › Conference contribution
}
TY - GEN
T1 - Accelerating advanced MRI reconstructions on GPUs
AU - Stone, Sam S.
AU - Haldar, Justin P.
AU - Tsao, Stephanie C.
AU - Hwu, Wen Mei W
AU - Liang, Zhi Pei
AU - Sutton, Bradley P.
PY - 2008/1/1
Y1 - 2008/1/1
N2 - Computational acceleration on graphics processing units (GPUs) can make advanced magnetic resonance imaging (MRI) reconstruction algorithms attractive in clinical settings, thereby improving the quality of MR images across a broad spectrum of applications. At present, MR imaging is often limited by high noise levels, significant imaging artifacts, and/or long data acquisition (scan) times. Advanced image reconstruction algorithms can mitigate these limitations and improve image quality by simultaneously operating on scan data acquired with arbitrary trajectories and incorporating additional information such as anatomical constraints. However, the improvements in image quality come at the expense of a considerable increase in computation. This paper describes the acceleration of an advanced reconstruction algorithm on NVIDIA's Quadro FX 5600. Optimizations such as register allocating the voxel data, tiling the scan data, and storing the scan data in the Quadro's constant memory dramatically reduce the reconstruction's required bandwidth to off-chip memory. The Quadro's special functional units provide substantial acceleration of the trigonometric computations in the algorithm's inner loops, and experimentally-tuned code transformations increase the reconstruction's performance by an additional 20%. The reconstruction of a 3D image with 1283 voxels ultimately achieves 150 GFLOPS and requires less than two minutes on the Quadro, while reconstruction on a quadcore CPU is thirteen times slower. Furthermore, relative to the true image, the error exhibited by the advanced reconstruction is only 12%, while conventional reconstruction techniques incur error of 42%. In short, the acceleration afforded by the GPU greatly increases the appeal of the advanced reconstruction for clinical MRI applications.
AB - Computational acceleration on graphics processing units (GPUs) can make advanced magnetic resonance imaging (MRI) reconstruction algorithms attractive in clinical settings, thereby improving the quality of MR images across a broad spectrum of applications. At present, MR imaging is often limited by high noise levels, significant imaging artifacts, and/or long data acquisition (scan) times. Advanced image reconstruction algorithms can mitigate these limitations and improve image quality by simultaneously operating on scan data acquired with arbitrary trajectories and incorporating additional information such as anatomical constraints. However, the improvements in image quality come at the expense of a considerable increase in computation. This paper describes the acceleration of an advanced reconstruction algorithm on NVIDIA's Quadro FX 5600. Optimizations such as register allocating the voxel data, tiling the scan data, and storing the scan data in the Quadro's constant memory dramatically reduce the reconstruction's required bandwidth to off-chip memory. The Quadro's special functional units provide substantial acceleration of the trigonometric computations in the algorithm's inner loops, and experimentally-tuned code transformations increase the reconstruction's performance by an additional 20%. The reconstruction of a 3D image with 1283 voxels ultimately achieves 150 GFLOPS and requires less than two minutes on the Quadro, while reconstruction on a quadcore CPU is thirteen times slower. Furthermore, relative to the true image, the error exhibited by the advanced reconstruction is only 12%, while conventional reconstruction techniques incur error of 42%. In short, the acceleration afforded by the GPU greatly increases the appeal of the advanced reconstruction for clinical MRI applications.
KW - CUDA
KW - GPGPU
KW - GPU computing
KW - MRI
KW - Reconstruction
UR - http://www.scopus.com/inward/record.url?scp=56749139615&partnerID=8YFLogxK
UR - http://www.scopus.com/inward/citedby.url?scp=56749139615&partnerID=8YFLogxK
U2 - 10.1145/1366230.1366276
DO - 10.1145/1366230.1366276
M3 - Conference contribution
AN - SCOPUS:56749139615
SN - 9781605580777
T3 - Conference on Computing Frontiers - Proceedings of the 2008 Conference on Computing Frontiers, CF'08
SP - 261
EP - 272
BT - Conference on Computing Frontiers - Proceedings of the 2008 Conference on Computing Frontiers, CF'08
PB - Association for Computing Machinery
ER -