High-performance 3D Compressive Sensing MRI reconstruction

Daehyun Kim, Joshua D Trazasko, Mikhail Smelyanskiy, Clifton R Haider, Armando Manduca, Pradeep Dubey

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

11 Citations (Scopus)

Abstract

Compressive Sensing (CS) is a nascent sampling and reconstruction paradigm that describes how sparse or compressible signals can be accurately approximated using many fewer samples than traditionally believed. In magnetic resonance imaging (MRI), where scan duration is directly proportional to the number of acquired samples, CS has the potential to dramatically decrease scan time. However, the computationally expensive nature of CS reconstructions has so far precluded their use in routine clinical practice - instead, more-easily generated but lower-quality images continue to be used. We investigate the development and optimization of a proven inexact quasi-Newton CS reconstruction algorithm on several modern parallel architectures, including CPUs, GPUs, and Intel's Many Integrated Core (MIC) architecture. Our (optimized) baseline implementation on a quad-core Core i7 is able to reconstruct a 256x160x80 volume of the neurovasculature from an 8-channel, 10x undersampled data set within 56 seconds, which is already a significant improvement over existing implementations. The latest six-core Core i7 reduces the reconstruction time further to 32 seconds. Moreover, we show that the CS algorithm benefits from modern throughput-oriented architectures. Specifically, our CUDA-base implementation on NVIDIA GTX480 reconstructs the same dataset in 16 seconds, while Intel's Knights Ferry (KNF) of the MIC architecture even reduces the time to 12 seconds. Such level of performance allows the neurovascular dataset to be reconstructed within a clinically viable time.

Original languageEnglish (US)
Title of host publication2010 Annual International Conference of the IEEE Engineering in Medicine and Biology Society, EMBC'10
Pages3321-3324
Number of pages4
DOIs
StatePublished - 2010
Event2010 32nd Annual International Conference of the IEEE Engineering in Medicine and Biology Society, EMBC'10 - Buenos Aires, Argentina
Duration: Aug 31 2010Sep 4 2010

Other

Other2010 32nd Annual International Conference of the IEEE Engineering in Medicine and Biology Society, EMBC'10
CountryArgentina
CityBuenos Aires
Period8/31/109/4/10

Fingerprint

Magnetic resonance
Imaging techniques
Parallel architectures
Image quality
Program processors
Throughput
Sampling
Graphics processing unit

Keywords

  • Angiography
  • Compressive Sensing
  • GPU
  • Many Core
  • MRI
  • Reconstruction

ASJC Scopus subject areas

  • Biomedical Engineering

Cite this

Kim, D., Trazasko, J. D., Smelyanskiy, M., Haider, C. R., Manduca, A., & Dubey, P. (2010). High-performance 3D Compressive Sensing MRI reconstruction. In 2010 Annual International Conference of the IEEE Engineering in Medicine and Biology Society, EMBC'10 (pp. 3321-3324). [5627493] https://doi.org/10.1109/IEMBS.2010.5627493

High-performance 3D Compressive Sensing MRI reconstruction. / Kim, Daehyun; Trazasko, Joshua D; Smelyanskiy, Mikhail; Haider, Clifton R; Manduca, Armando; Dubey, Pradeep.

2010 Annual International Conference of the IEEE Engineering in Medicine and Biology Society, EMBC'10. 2010. p. 3321-3324 5627493.

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

Kim, D, Trazasko, JD, Smelyanskiy, M, Haider, CR, Manduca, A & Dubey, P 2010, High-performance 3D Compressive Sensing MRI reconstruction. in 2010 Annual International Conference of the IEEE Engineering in Medicine and Biology Society, EMBC'10., 5627493, pp. 3321-3324, 2010 32nd Annual International Conference of the IEEE Engineering in Medicine and Biology Society, EMBC'10, Buenos Aires, Argentina, 8/31/10. https://doi.org/10.1109/IEMBS.2010.5627493
Kim D, Trazasko JD, Smelyanskiy M, Haider CR, Manduca A, Dubey P. High-performance 3D Compressive Sensing MRI reconstruction. In 2010 Annual International Conference of the IEEE Engineering in Medicine and Biology Society, EMBC'10. 2010. p. 3321-3324. 5627493 https://doi.org/10.1109/IEMBS.2010.5627493
Kim, Daehyun ; Trazasko, Joshua D ; Smelyanskiy, Mikhail ; Haider, Clifton R ; Manduca, Armando ; Dubey, Pradeep. / High-performance 3D Compressive Sensing MRI reconstruction. 2010 Annual International Conference of the IEEE Engineering in Medicine and Biology Society, EMBC'10. 2010. pp. 3321-3324
@inproceedings{728f41279b254bc6948102c2d390ac7b,
title = "High-performance 3D Compressive Sensing MRI reconstruction",
abstract = "Compressive Sensing (CS) is a nascent sampling and reconstruction paradigm that describes how sparse or compressible signals can be accurately approximated using many fewer samples than traditionally believed. In magnetic resonance imaging (MRI), where scan duration is directly proportional to the number of acquired samples, CS has the potential to dramatically decrease scan time. However, the computationally expensive nature of CS reconstructions has so far precluded their use in routine clinical practice - instead, more-easily generated but lower-quality images continue to be used. We investigate the development and optimization of a proven inexact quasi-Newton CS reconstruction algorithm on several modern parallel architectures, including CPUs, GPUs, and Intel's Many Integrated Core (MIC) architecture. Our (optimized) baseline implementation on a quad-core Core i7 is able to reconstruct a 256x160x80 volume of the neurovasculature from an 8-channel, 10x undersampled data set within 56 seconds, which is already a significant improvement over existing implementations. The latest six-core Core i7 reduces the reconstruction time further to 32 seconds. Moreover, we show that the CS algorithm benefits from modern throughput-oriented architectures. Specifically, our CUDA-base implementation on NVIDIA GTX480 reconstructs the same dataset in 16 seconds, while Intel's Knights Ferry (KNF) of the MIC architecture even reduces the time to 12 seconds. Such level of performance allows the neurovascular dataset to be reconstructed within a clinically viable time.",
keywords = "Angiography, Compressive Sensing, GPU, Many Core, MRI, Reconstruction",
author = "Daehyun Kim and Trazasko, {Joshua D} and Mikhail Smelyanskiy and Haider, {Clifton R} and Armando Manduca and Pradeep Dubey",
year = "2010",
doi = "10.1109/IEMBS.2010.5627493",
language = "English (US)",
isbn = "9781424441235",
pages = "3321--3324",
booktitle = "2010 Annual International Conference of the IEEE Engineering in Medicine and Biology Society, EMBC'10",

}

TY - GEN

T1 - High-performance 3D Compressive Sensing MRI reconstruction

AU - Kim, Daehyun

AU - Trazasko, Joshua D

AU - Smelyanskiy, Mikhail

AU - Haider, Clifton R

AU - Manduca, Armando

AU - Dubey, Pradeep

PY - 2010

Y1 - 2010

N2 - Compressive Sensing (CS) is a nascent sampling and reconstruction paradigm that describes how sparse or compressible signals can be accurately approximated using many fewer samples than traditionally believed. In magnetic resonance imaging (MRI), where scan duration is directly proportional to the number of acquired samples, CS has the potential to dramatically decrease scan time. However, the computationally expensive nature of CS reconstructions has so far precluded their use in routine clinical practice - instead, more-easily generated but lower-quality images continue to be used. We investigate the development and optimization of a proven inexact quasi-Newton CS reconstruction algorithm on several modern parallel architectures, including CPUs, GPUs, and Intel's Many Integrated Core (MIC) architecture. Our (optimized) baseline implementation on a quad-core Core i7 is able to reconstruct a 256x160x80 volume of the neurovasculature from an 8-channel, 10x undersampled data set within 56 seconds, which is already a significant improvement over existing implementations. The latest six-core Core i7 reduces the reconstruction time further to 32 seconds. Moreover, we show that the CS algorithm benefits from modern throughput-oriented architectures. Specifically, our CUDA-base implementation on NVIDIA GTX480 reconstructs the same dataset in 16 seconds, while Intel's Knights Ferry (KNF) of the MIC architecture even reduces the time to 12 seconds. Such level of performance allows the neurovascular dataset to be reconstructed within a clinically viable time.

AB - Compressive Sensing (CS) is a nascent sampling and reconstruction paradigm that describes how sparse or compressible signals can be accurately approximated using many fewer samples than traditionally believed. In magnetic resonance imaging (MRI), where scan duration is directly proportional to the number of acquired samples, CS has the potential to dramatically decrease scan time. However, the computationally expensive nature of CS reconstructions has so far precluded their use in routine clinical practice - instead, more-easily generated but lower-quality images continue to be used. We investigate the development and optimization of a proven inexact quasi-Newton CS reconstruction algorithm on several modern parallel architectures, including CPUs, GPUs, and Intel's Many Integrated Core (MIC) architecture. Our (optimized) baseline implementation on a quad-core Core i7 is able to reconstruct a 256x160x80 volume of the neurovasculature from an 8-channel, 10x undersampled data set within 56 seconds, which is already a significant improvement over existing implementations. The latest six-core Core i7 reduces the reconstruction time further to 32 seconds. Moreover, we show that the CS algorithm benefits from modern throughput-oriented architectures. Specifically, our CUDA-base implementation on NVIDIA GTX480 reconstructs the same dataset in 16 seconds, while Intel's Knights Ferry (KNF) of the MIC architecture even reduces the time to 12 seconds. Such level of performance allows the neurovascular dataset to be reconstructed within a clinically viable time.

KW - Angiography

KW - Compressive Sensing

KW - GPU

KW - Many Core

KW - MRI

KW - Reconstruction

UR - http://www.scopus.com/inward/record.url?scp=78650830903&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=78650830903&partnerID=8YFLogxK

U2 - 10.1109/IEMBS.2010.5627493

DO - 10.1109/IEMBS.2010.5627493

M3 - Conference contribution

SN - 9781424441235

SP - 3321

EP - 3324

BT - 2010 Annual International Conference of the IEEE Engineering in Medicine and Biology Society, EMBC'10

ER -