SRAKI-RNN: Accelerated MRI with scan-specific recurrent neural networks using densely connected blocks

Seyed Amir Hossein Hosseini; Chi Zhang; Kâmil Ugurbil; Steen Moeller; Mehmet Akcąkaya

doi:10.1117/12.2527949

SRAKI-RNN: Accelerated MRI with scan-specific recurrent neural networks using densely connected blocks

Seyed Amir Hossein Hosseini, Chi Zhang, Kâmil Ugurbil, Steen Moeller, Mehmet Akcąkaya

Research output: Chapter in Book/Report/Conference proceeding › Conference contribution

8 Scopus citations

Abstract

This study aims to improve upon Self-consistent Robust Artificial-neural-networks for k-space Interpolation (sRAKI), which is a deep learning-based parallel imaging technique for accelerated MRI reconstruction. The proposed technique, called sRAKI-RNN, combines the calibration and reconstruction phases of sRAKI into a single step that jointly learns the self-consistency rule and performs iterative reconstruction using recurrent neural networks (RNN). Similar to sRAKI, sRAKI-RNN supports arbitrary undersampling patterns and is a databasefree technique that is trained on autocalibrating signal (ACS) data from the same scan. Densely connected blocks are used in each iteration of the RNN to improve the convergence during the learning phase. sRAKI-RNN was evaluated on targeted right coronary artery (RCA) MRI. The results indicate that sRAKI-RNN further improves the noise resilience of sRAKI in a shorter running time and also considerably outperforms its linear counterpart, SPIRiT, in suppressing reconstruction noise.

Original language	English (US)
Title of host publication	Wavelets and Sparsity XVIII
Editors	Dimitri Van De Ville, Dimitri Van De Ville, Manos Papadakis, Yue M. Lu
Publisher	SPIE
ISBN (Electronic)	9781510629691
DOIs	https://doi.org/10.1117/12.2527949
State	Published - 2019
Event	Wavelets and Sparsity XVIII 2019 - San Diego, United States Duration: Aug 13 2019 → Aug 15 2019

Publication series

Name	Proceedings of SPIE - The International Society for Optical Engineering
Volume	11138
ISSN (Print)	0277-786X
ISSN (Electronic)	1996-756X

Conference

Conference	Wavelets and Sparsity XVIII 2019
Country/Territory	United States
City	San Diego
Period	8/13/19 → 8/15/19

Bibliographical note

Funding Information:
This work was supported by NIH, Grant numbers: NSF, Grant number: CAREER CCF-1651825

Publisher Copyright:
© 2019 SPIE.

Keywords

Accelerated MRI
coronary MRI
deep learning
dense neural networks
parallel imaging
recurrent neural networks

Publisher link

10.1117/12.2527949

Find It

Find It at U of M Libraries

Cite this

Hosseini, S. A. H., Zhang, C., Ugurbil, K., Moeller, S., & Akcąkaya, M. (2019). SRAKI-RNN: Accelerated MRI with scan-specific recurrent neural networks using densely connected blocks. In D. Van De Ville, D. Van De Ville, M. Papadakis, & Y. M. Lu (Eds.), Wavelets and Sparsity XVIII [111381B] (Proceedings of SPIE - The International Society for Optical Engineering; Vol. 11138). SPIE. https://doi.org/10.1117/12.2527949

SRAKI-RNN : Accelerated MRI with scan-specific recurrent neural networks using densely connected blocks. / Hosseini, Seyed Amir Hossein; Zhang, Chi; Ugurbil, Kâmil et al.

Wavelets and Sparsity XVIII. ed. / Dimitri Van De Ville; Dimitri Van De Ville; Manos Papadakis; Yue M. Lu. SPIE, 2019. 111381B (Proceedings of SPIE - The International Society for Optical Engineering; Vol. 11138).

Research output: Chapter in Book/Report/Conference proceeding › Conference contribution

Hosseini, SAH, Zhang, C, Ugurbil, K , Moeller, S & Akcąkaya, M 2019, SRAKI-RNN: Accelerated MRI with scan-specific recurrent neural networks using densely connected blocks. in D Van De Ville, D Van De Ville, M Papadakis & YM Lu (eds), Wavelets and Sparsity XVIII., 111381B, Proceedings of SPIE - The International Society for Optical Engineering, vol. 11138, SPIE, Wavelets and Sparsity XVIII 2019, San Diego, United States, 8/13/19. https://doi.org/10.1117/12.2527949

Hosseini SAH, Zhang C, Ugurbil K , Moeller S , Akcąkaya M. SRAKI-RNN: Accelerated MRI with scan-specific recurrent neural networks using densely connected blocks. In Van De Ville D, Van De Ville D, Papadakis M, Lu YM, editors, Wavelets and Sparsity XVIII. SPIE. 2019. 111381B. (Proceedings of SPIE - The International Society for Optical Engineering). doi: 10.1117/12.2527949

Hosseini, Seyed Amir Hossein ; Zhang, Chi ; Ugurbil, Kâmil et al. / SRAKI-RNN : Accelerated MRI with scan-specific recurrent neural networks using densely connected blocks. Wavelets and Sparsity XVIII. editor / Dimitri Van De Ville ; Dimitri Van De Ville ; Manos Papadakis ; Yue M. Lu. SPIE, 2019. (Proceedings of SPIE - The International Society for Optical Engineering).

@inproceedings{fb2fa1d0f2384d9f9e4e61dd02da1073,

title = "SRAKI-RNN: Accelerated MRI with scan-specific recurrent neural networks using densely connected blocks",

abstract = "This study aims to improve upon Self-consistent Robust Artificial-neural-networks for k-space Interpolation (sRAKI), which is a deep learning-based parallel imaging technique for accelerated MRI reconstruction. The proposed technique, called sRAKI-RNN, combines the calibration and reconstruction phases of sRAKI into a single step that jointly learns the self-consistency rule and performs iterative reconstruction using recurrent neural networks (RNN). Similar to sRAKI, sRAKI-RNN supports arbitrary undersampling patterns and is a databasefree technique that is trained on autocalibrating signal (ACS) data from the same scan. Densely connected blocks are used in each iteration of the RNN to improve the convergence during the learning phase. sRAKI-RNN was evaluated on targeted right coronary artery (RCA) MRI. The results indicate that sRAKI-RNN further improves the noise resilience of sRAKI in a shorter running time and also considerably outperforms its linear counterpart, SPIRiT, in suppressing reconstruction noise.",

keywords = "Accelerated MRI, coronary MRI, deep learning, dense neural networks, parallel imaging, recurrent neural networks",

author = "Hosseini, {Seyed Amir Hossein} and Chi Zhang and K{\^a}mil Ugurbil and Steen Moeller and Mehmet Akc{\c a}kaya",

note = "Funding Information: This work was supported by NIH, Grant numbers: NSF, Grant number: CAREER CCF-1651825 Publisher Copyright: {\textcopyright} 2019 SPIE.; Wavelets and Sparsity XVIII 2019 ; Conference date: 13-08-2019 Through 15-08-2019",

year = "2019",

doi = "10.1117/12.2527949",

language = "English (US)",

series = "Proceedings of SPIE - The International Society for Optical Engineering",

publisher = "SPIE",

editor = "{Van De Ville}, Dimitri and {Van De Ville}, Dimitri and Manos Papadakis and Lu, {Yue M.}",

booktitle = "Wavelets and Sparsity XVIII",

address = "United States",

}

TY - GEN

T1 - SRAKI-RNN

T2 - Wavelets and Sparsity XVIII 2019

AU - Hosseini, Seyed Amir Hossein

AU - Zhang, Chi

AU - Ugurbil, Kâmil

AU - Moeller, Steen

AU - Akcąkaya, Mehmet

PY - 2019

Y1 - 2019

N2 - This study aims to improve upon Self-consistent Robust Artificial-neural-networks for k-space Interpolation (sRAKI), which is a deep learning-based parallel imaging technique for accelerated MRI reconstruction. The proposed technique, called sRAKI-RNN, combines the calibration and reconstruction phases of sRAKI into a single step that jointly learns the self-consistency rule and performs iterative reconstruction using recurrent neural networks (RNN). Similar to sRAKI, sRAKI-RNN supports arbitrary undersampling patterns and is a databasefree technique that is trained on autocalibrating signal (ACS) data from the same scan. Densely connected blocks are used in each iteration of the RNN to improve the convergence during the learning phase. sRAKI-RNN was evaluated on targeted right coronary artery (RCA) MRI. The results indicate that sRAKI-RNN further improves the noise resilience of sRAKI in a shorter running time and also considerably outperforms its linear counterpart, SPIRiT, in suppressing reconstruction noise.

AB - This study aims to improve upon Self-consistent Robust Artificial-neural-networks for k-space Interpolation (sRAKI), which is a deep learning-based parallel imaging technique for accelerated MRI reconstruction. The proposed technique, called sRAKI-RNN, combines the calibration and reconstruction phases of sRAKI into a single step that jointly learns the self-consistency rule and performs iterative reconstruction using recurrent neural networks (RNN). Similar to sRAKI, sRAKI-RNN supports arbitrary undersampling patterns and is a databasefree technique that is trained on autocalibrating signal (ACS) data from the same scan. Densely connected blocks are used in each iteration of the RNN to improve the convergence during the learning phase. sRAKI-RNN was evaluated on targeted right coronary artery (RCA) MRI. The results indicate that sRAKI-RNN further improves the noise resilience of sRAKI in a shorter running time and also considerably outperforms its linear counterpart, SPIRiT, in suppressing reconstruction noise.

KW - Accelerated MRI

KW - coronary MRI

KW - deep learning

KW - dense neural networks

KW - parallel imaging

KW - recurrent neural networks

UR - http://www.scopus.com/inward/record.url?scp=85077131478&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=85077131478&partnerID=8YFLogxK

U2 - 10.1117/12.2527949

DO - 10.1117/12.2527949

M3 - Conference contribution

AN - SCOPUS:85077131478

T3 - Proceedings of SPIE - The International Society for Optical Engineering

BT - Wavelets and Sparsity XVIII

A2 - Van De Ville, Dimitri

A2 - Papadakis, Manos

A2 - Lu, Yue M.

PB - SPIE

Y2 - 13 August 2019 through 15 August 2019

ER -

SRAKI-RNN: Accelerated MRI with scan-specific recurrent neural networks using densely connected blocks

Abstract

Publication series

Conference

Bibliographical note

Keywords

Publisher link

Find It

Other files and links

Fingerprint

Cite this