A multi-scale variational neural network for accelerating motion-compensated whole-heart 3D coronary MR angiography

Niccolo Fuin; Aurelien Bustin; Thomas Küstner; Ilkay Oksuz; James Clough; Andrew P. King; Julia A. Schnabel; René M. Botnar; Claudia Prieto

doi:10.1016/j.mri.2020.04.007

A multi-scale variational neural network for accelerating motion-compensated whole-heart 3D coronary MR angiography

Niccolo Fuin^*, Aurelien Bustin, Thomas Küstner, Ilkay Oksuz, James Clough, Andrew P. King, Julia A. Schnabel, René M. Botnar, Claudia Prieto

^*Corresponding author for this work

Department of Computer Engineering

Research output: Contribution to journal › Article › peer-review

37 Citations (Scopus)

Abstract

Purpose: To enable fast reconstruction of undersampled motion-compensated whole-heart 3D coronary magnetic resonance angiography (CMRA) by learning a multi-scale variational neural network (MS-VNN) which allows the acquisition of high-quality 1.2 × 1.2 × 1.2 mm isotropic volumes in a short and predictable scan time. Methods: Eighteen healthy subjects and one patient underwent free-breathing 3D CMRA acquisition with variable density spiral-like Cartesian sampling, combined with 2D image navigators for translational motion estimation/compensation. The proposed MS-VNN learns two sets of kernels and activation functions for the magnitude and phase images of the complex-valued data. For the magnitude, a multi-scale approach is applied to better capture the small calibre of the coronaries. Ten subjects were considered for training and validation. Prospectively undersampled motion-compensated data with 5-fold and 9-fold accelerations, from the remaining 9 subjects, were used to evaluate the framework. The proposed approach was compared to Wavelet-based compressed-sensing (CS), conventional VNN, and to an additional fully-sampled (FS) scan. Results: The average acquisition time (m:s) was 4:11 for 5-fold, 2:34 for 9-fold acceleration and 18:55 for fully-sampled. Reconstruction time with the proposed MS-VNN was ~14 s. The proposed MS-VNN achieves higher image quality than CS and VNN reconstructions, with quantitative right coronary artery sharpness (CS:43.0%, VNN:43.9%, MS-VNN:47.0%, FS:50.67%) and vessel length (CS:7.4 cm, VNN:7.7 cm, MS-VNN:8.8 cm, FS:9.1 cm) comparable to the FS scan. Conclusion: The proposed MS-VNN enables 5-fold and 9-fold undersampled CMRA acquisitions with comparable image quality that the corresponding fully-sampled scan. The proposed framework achieves extremely fast reconstruction time and does not require tuning of regularization parameters, offering easy integration into clinical workflow.

Original language	English
Pages (from-to)	155-167
Number of pages	13
Journal	Magnetic Resonance Imaging
Volume	70
DOIs	https://doi.org/10.1016/j.mri.2020.04.007
Publication status	Published - Jul 2020

Bibliographical note

Publisher Copyright:
© 2020

Funding

The authors acknowledge financial support from: (1) Engineering and Physical Sciences Research Council (EPSRC), UK: EP/P001009/1, EP/P032311/1, EPSRC EP/P007619, (2) Wellcome EPSRC Centre for Medical Engineering , UK: NS/A000049/1 , and (3) the Department of Health via the National Institute for Health Research (NIHR), UK: comprehensive Biomedical Research Centre award to Guy's & St Thomas' NHS Foundation Trust in partnership with King's College London and King's College Hospital NHS Foundation Trust. We gratefully acknowledge the support of NVIDIA Corporation with the donation of the Titan V GPU used for this research. The views expressed are those of the authors and not necessarily those of the NHS, the NIHR or the Department of Health. The authors acknowledge financial support from: (1) Engineering and Physical Sciences Research Council (EPSRC), UK: EP/P001009/1, EP/P032311/1, EPSRC EP/P007619, (2) Wellcome EPSRC Centre for Medical Engineering, UK:NS/A000049/1, and (3) the Department of Health via the National Institute for Health Research (NIHR), UK: comprehensive Biomedical Research Centre award to Guy's & St Thomas' NHS Foundation Trust in partnership with King's College London and King's College Hospital NHS Foundation Trust. We gratefully acknowledge the support of NVIDIA Corporation with the donation of the Titan V GPU used for this research. The views expressed are those of the authors and not necessarily those of the NHS, the NIHR or the Department of Health.

Funders	Funder number
Wellcome EPSRC	NS/A000049/1
NVIDIA
King's College Hospital NHS Foundation Trust
NIHR Biomedical Research Centre, Royal Marsden NHS Foundation Trust/Institute of Cancer Research
Engineering and Physical Sciences Research Council	EP/P032311/1, EP/P001009/1, EP/P007619
National Institute for Health Research
Department of Health and Social Care
King's College London
Department of Health, Australian Government

Keywords

Cardiac MRI
Coronary imaging
Fast imaging
Undersampling
Variational neural network

UN SDGs

This output contributes to the following UN Sustainable Development Goals (SDGs)

Access to Document

10.1016/j.mri.2020.04.007

Cite this

@article{ef80fc964c294ab28d0a40cc64a225bb,

title = "A multi-scale variational neural network for accelerating motion-compensated whole-heart 3D coronary MR angiography",

abstract = "Purpose: To enable fast reconstruction of undersampled motion-compensated whole-heart 3D coronary magnetic resonance angiography (CMRA) by learning a multi-scale variational neural network (MS-VNN) which allows the acquisition of high-quality 1.2 × 1.2 × 1.2 mm isotropic volumes in a short and predictable scan time. Methods: Eighteen healthy subjects and one patient underwent free-breathing 3D CMRA acquisition with variable density spiral-like Cartesian sampling, combined with 2D image navigators for translational motion estimation/compensation. The proposed MS-VNN learns two sets of kernels and activation functions for the magnitude and phase images of the complex-valued data. For the magnitude, a multi-scale approach is applied to better capture the small calibre of the coronaries. Ten subjects were considered for training and validation. Prospectively undersampled motion-compensated data with 5-fold and 9-fold accelerations, from the remaining 9 subjects, were used to evaluate the framework. The proposed approach was compared to Wavelet-based compressed-sensing (CS), conventional VNN, and to an additional fully-sampled (FS) scan. Results: The average acquisition time (m:s) was 4:11 for 5-fold, 2:34 for 9-fold acceleration and 18:55 for fully-sampled. Reconstruction time with the proposed MS-VNN was ~14 s. The proposed MS-VNN achieves higher image quality than CS and VNN reconstructions, with quantitative right coronary artery sharpness (CS:43.0%, VNN:43.9%, MS-VNN:47.0%, FS:50.67%) and vessel length (CS:7.4 cm, VNN:7.7 cm, MS-VNN:8.8 cm, FS:9.1 cm) comparable to the FS scan. Conclusion: The proposed MS-VNN enables 5-fold and 9-fold undersampled CMRA acquisitions with comparable image quality that the corresponding fully-sampled scan. The proposed framework achieves extremely fast reconstruction time and does not require tuning of regularization parameters, offering easy integration into clinical workflow.",

keywords = "Cardiac MRI, Coronary imaging, Fast imaging, Undersampling, Variational neural network",

author = "Niccolo Fuin and Aurelien Bustin and Thomas K{\"u}stner and Ilkay Oksuz and James Clough and King, {Andrew P.} and Schnabel, {Julia A.} and Botnar, {Ren{\'e} M.} and Claudia Prieto",

note = "Publisher Copyright: {\textcopyright} 2020",

year = "2020",

month = jul,

doi = "10.1016/j.mri.2020.04.007",

language = "English",

volume = "70",

pages = "155--167",

journal = "Magnetic Resonance Imaging",

issn = "0730-725X",

publisher = "Elsevier Inc.",

}

TY - JOUR

T1 - A multi-scale variational neural network for accelerating motion-compensated whole-heart 3D coronary MR angiography

AU - Fuin, Niccolo

AU - Bustin, Aurelien

AU - Küstner, Thomas

AU - Oksuz, Ilkay

AU - Clough, James

AU - King, Andrew P.

AU - Schnabel, Julia A.

AU - Botnar, René M.

AU - Prieto, Claudia

PY - 2020/7

Y1 - 2020/7

N2 - Purpose: To enable fast reconstruction of undersampled motion-compensated whole-heart 3D coronary magnetic resonance angiography (CMRA) by learning a multi-scale variational neural network (MS-VNN) which allows the acquisition of high-quality 1.2 × 1.2 × 1.2 mm isotropic volumes in a short and predictable scan time. Methods: Eighteen healthy subjects and one patient underwent free-breathing 3D CMRA acquisition with variable density spiral-like Cartesian sampling, combined with 2D image navigators for translational motion estimation/compensation. The proposed MS-VNN learns two sets of kernels and activation functions for the magnitude and phase images of the complex-valued data. For the magnitude, a multi-scale approach is applied to better capture the small calibre of the coronaries. Ten subjects were considered for training and validation. Prospectively undersampled motion-compensated data with 5-fold and 9-fold accelerations, from the remaining 9 subjects, were used to evaluate the framework. The proposed approach was compared to Wavelet-based compressed-sensing (CS), conventional VNN, and to an additional fully-sampled (FS) scan. Results: The average acquisition time (m:s) was 4:11 for 5-fold, 2:34 for 9-fold acceleration and 18:55 for fully-sampled. Reconstruction time with the proposed MS-VNN was ~14 s. The proposed MS-VNN achieves higher image quality than CS and VNN reconstructions, with quantitative right coronary artery sharpness (CS:43.0%, VNN:43.9%, MS-VNN:47.0%, FS:50.67%) and vessel length (CS:7.4 cm, VNN:7.7 cm, MS-VNN:8.8 cm, FS:9.1 cm) comparable to the FS scan. Conclusion: The proposed MS-VNN enables 5-fold and 9-fold undersampled CMRA acquisitions with comparable image quality that the corresponding fully-sampled scan. The proposed framework achieves extremely fast reconstruction time and does not require tuning of regularization parameters, offering easy integration into clinical workflow.

AB - Purpose: To enable fast reconstruction of undersampled motion-compensated whole-heart 3D coronary magnetic resonance angiography (CMRA) by learning a multi-scale variational neural network (MS-VNN) which allows the acquisition of high-quality 1.2 × 1.2 × 1.2 mm isotropic volumes in a short and predictable scan time. Methods: Eighteen healthy subjects and one patient underwent free-breathing 3D CMRA acquisition with variable density spiral-like Cartesian sampling, combined with 2D image navigators for translational motion estimation/compensation. The proposed MS-VNN learns two sets of kernels and activation functions for the magnitude and phase images of the complex-valued data. For the magnitude, a multi-scale approach is applied to better capture the small calibre of the coronaries. Ten subjects were considered for training and validation. Prospectively undersampled motion-compensated data with 5-fold and 9-fold accelerations, from the remaining 9 subjects, were used to evaluate the framework. The proposed approach was compared to Wavelet-based compressed-sensing (CS), conventional VNN, and to an additional fully-sampled (FS) scan. Results: The average acquisition time (m:s) was 4:11 for 5-fold, 2:34 for 9-fold acceleration and 18:55 for fully-sampled. Reconstruction time with the proposed MS-VNN was ~14 s. The proposed MS-VNN achieves higher image quality than CS and VNN reconstructions, with quantitative right coronary artery sharpness (CS:43.0%, VNN:43.9%, MS-VNN:47.0%, FS:50.67%) and vessel length (CS:7.4 cm, VNN:7.7 cm, MS-VNN:8.8 cm, FS:9.1 cm) comparable to the FS scan. Conclusion: The proposed MS-VNN enables 5-fold and 9-fold undersampled CMRA acquisitions with comparable image quality that the corresponding fully-sampled scan. The proposed framework achieves extremely fast reconstruction time and does not require tuning of regularization parameters, offering easy integration into clinical workflow.

KW - Cardiac MRI

KW - Coronary imaging

KW - Fast imaging

KW - Undersampling

KW - Variational neural network

UR - http://www.scopus.com/inward/record.url?scp=85084453689&partnerID=8YFLogxK

U2 - 10.1016/j.mri.2020.04.007

DO - 10.1016/j.mri.2020.04.007

M3 - Article

C2 - 32353528

AN - SCOPUS:85084453689

SN - 0730-725X

VL - 70

SP - 155

EP - 167

JO - Magnetic Resonance Imaging

JF - Magnetic Resonance Imaging

ER -

A multi-scale variational neural network for accelerating motion-compensated whole-heart 3D coronary MR angiography

Abstract

Bibliographical note

Funding

Keywords

UN SDGs

Access to Document

Other files and links

Fingerprint

Cite this