Nonlinear Equivariant Imaging: Learning Multi-Parametric Tissue Mapping without Ground Truth for Compressive Quantitative MRI

Ketan Fatania; Kwai Y. Chau; Carolin M. Pirkl; Marion I. Menzel; Mohammad Golbabaee

doi:10.1109/ISBI53787.2023.10230440

Nonlinear Equivariant Imaging: Learning Multi-Parametric Tissue Mapping without Ground Truth for Compressive Quantitative MRI

Ketan Fatania, Kwai Y. Chau, Carolin M. Pirkl, Marion I. Menzel, Mohammad Golbabaee

Research output: Chapter or section in a book/report/conference proceeding › Chapter in a published conference proceeding

3 Citations (SciVal)

76 Downloads (Pure)

Abstract

Current state-of-the-art reconstruction for quantitative tissue maps from fast, compressive, Magnetic Resonance Fingerprinting (MRF), use supervised deep learning, with the drawback of requiring high-fidelity ground truth tissue map training data which is limited. This paper proposes NonLinear Equivariant Imaging for MRF (NLEIMRF), a self-supervised learning approach to eliminate the need for ground truth for deep MRF image reconstruction. NLEI-MRF extends the recent Equivariant Imaging framework to the MRF nonlinear inverse problem. Only compressed-sampled MRF scans are used for training. NLEI-MRF learns tissue mapping using spatiotemporal priors: spatial priors are obtained from the invariance of MRF data to a group of geometric image transformations, while temporal priors are obtained from a nonlinear Bloch response model approximated by a pre-trained neural network. Tested retrospectively on two acquisition settings, we observe that NLEI-MRF closely approaches the performance of supervised learning.

Original language	English
Title of host publication	2023 IEEE International Symposium on Biomedical Imaging, ISBI 2023
Place of Publication	U. S. A.
Publisher	IEEE
ISBN (Electronic)	9781665473583
DOIs	https://doi.org/10.1109/ISBI53787.2023.10230440
Publication status	Published - 1 Sept 2023
Event	20th IEEE International Symposium on Biomedical Imaging, ISBI 2023 - Cartagena, Colombia Duration: 18 Apr 2023 → 21 Apr 2023

Publication series

Name	Proceedings - International Symposium on Biomedical Imaging
Volume	2023-April
ISSN (Print)	1945-7928
ISSN (Electronic)	1945-8452

Conference

Conference	20th IEEE International Symposium on Biomedical Imaging, ISBI 2023
Country/Territory	Colombia
City	Cartagena
Period	18/04/23 → 21/04/23

Bibliographical note

Funding Information:
CMP and MIM are supported by the EU's Horizon 2020 (grant No. 952172). MG is supported by the EPSRC grant EP/X001091/1.

Funding

CMP and MIM are supported by the EU's Horizon 2020 (grant No. 952172). MG is supported by the EPSRC grant EP/X001091/1.

Keywords

Compressed Sensing
Equivariant Imaging
Inverse Problems
Magnetic Resonance Fingerprinting
Quantitative MRI
Self-Supervised Deep Learning

ASJC Scopus subject areas

Biomedical Engineering
Radiology Nuclear Medicine and imaging

Access to Document

10.1109/ISBI53787.2023.10230440

QMRI_MRF_NLEI_Reconstruction_2023Accepted author manuscript, 7.92 MBLicence: CC BY

Cite this

Fatania, K., Chau, K. Y., Pirkl, C. M., Menzel, M. I., & Golbabaee, M. (2023). Nonlinear Equivariant Imaging: Learning Multi-Parametric Tissue Mapping without Ground Truth for Compressive Quantitative MRI. In 2023 IEEE International Symposium on Biomedical Imaging, ISBI 2023 (Proceedings - International Symposium on Biomedical Imaging; Vol. 2023-April). IEEE. https://doi.org/10.1109/ISBI53787.2023.10230440

Nonlinear Equivariant Imaging: Learning Multi-Parametric Tissue Mapping without Ground Truth for Compressive Quantitative MRI. / Fatania, Ketan; Chau, Kwai Y.; Pirkl, Carolin M. et al.
2023 IEEE International Symposium on Biomedical Imaging, ISBI 2023. U. S. A.: IEEE, 2023. (Proceedings - International Symposium on Biomedical Imaging; Vol. 2023-April).

Research output: Chapter or section in a book/report/conference proceeding › Chapter in a published conference proceeding

Fatania, K, Chau, KY, Pirkl, CM, Menzel, MI & Golbabaee, M 2023, Nonlinear Equivariant Imaging: Learning Multi-Parametric Tissue Mapping without Ground Truth for Compressive Quantitative MRI. in 2023 IEEE International Symposium on Biomedical Imaging, ISBI 2023. Proceedings - International Symposium on Biomedical Imaging, vol. 2023-April, IEEE, U. S. A., 20th IEEE International Symposium on Biomedical Imaging, ISBI 2023, Cartagena, Colombia, 18/04/23. https://doi.org/10.1109/ISBI53787.2023.10230440

Fatania K, Chau KY, Pirkl CM, Menzel MI, Golbabaee M. Nonlinear Equivariant Imaging: Learning Multi-Parametric Tissue Mapping without Ground Truth for Compressive Quantitative MRI. In 2023 IEEE International Symposium on Biomedical Imaging, ISBI 2023. U. S. A.: IEEE. 2023. (Proceedings - International Symposium on Biomedical Imaging). doi: 10.1109/ISBI53787.2023.10230440

@inproceedings{656aea8862fc4b6db5101be50e0e8d6b,

title = "Nonlinear Equivariant Imaging: Learning Multi-Parametric Tissue Mapping without Ground Truth for Compressive Quantitative MRI",

abstract = "Current state-of-the-art reconstruction for quantitative tissue maps from fast, compressive, Magnetic Resonance Fingerprinting (MRF), use supervised deep learning, with the drawback of requiring high-fidelity ground truth tissue map training data which is limited. This paper proposes NonLinear Equivariant Imaging for MRF (NLEIMRF), a self-supervised learning approach to eliminate the need for ground truth for deep MRF image reconstruction. NLEI-MRF extends the recent Equivariant Imaging framework to the MRF nonlinear inverse problem. Only compressed-sampled MRF scans are used for training. NLEI-MRF learns tissue mapping using spatiotemporal priors: spatial priors are obtained from the invariance of MRF data to a group of geometric image transformations, while temporal priors are obtained from a nonlinear Bloch response model approximated by a pre-trained neural network. Tested retrospectively on two acquisition settings, we observe that NLEI-MRF closely approaches the performance of supervised learning.",

keywords = "Compressed Sensing, Equivariant Imaging, Inverse Problems, Magnetic Resonance Fingerprinting, Quantitative MRI, Self-Supervised Deep Learning",

author = "Ketan Fatania and Chau, \{Kwai Y.\} and Pirkl, \{Carolin M.\} and Menzel, \{Marion I.\} and Mohammad Golbabaee",

note = "Funding Information: CMP and MIM are supported by the EU's Horizon 2020 (grant No. 952172). MG is supported by the EPSRC grant EP/X001091/1. ; 20th IEEE International Symposium on Biomedical Imaging, ISBI 2023 ; Conference date: 18-04-2023 Through 21-04-2023",

year = "2023",

month = sep,

day = "1",

doi = "10.1109/ISBI53787.2023.10230440",

language = "English",

series = "Proceedings - International Symposium on Biomedical Imaging",

publisher = "IEEE",

booktitle = "2023 IEEE International Symposium on Biomedical Imaging, ISBI 2023",

address = "USA United States",

}

TY - GEN

T1 - Nonlinear Equivariant Imaging

T2 - 20th IEEE International Symposium on Biomedical Imaging, ISBI 2023

AU - Fatania, Ketan

AU - Chau, Kwai Y.

AU - Pirkl, Carolin M.

AU - Menzel, Marion I.

AU - Golbabaee, Mohammad

N1 - Funding Information: CMP and MIM are supported by the EU's Horizon 2020 (grant No. 952172). MG is supported by the EPSRC grant EP/X001091/1.

PY - 2023/9/1

Y1 - 2023/9/1

N2 - Current state-of-the-art reconstruction for quantitative tissue maps from fast, compressive, Magnetic Resonance Fingerprinting (MRF), use supervised deep learning, with the drawback of requiring high-fidelity ground truth tissue map training data which is limited. This paper proposes NonLinear Equivariant Imaging for MRF (NLEIMRF), a self-supervised learning approach to eliminate the need for ground truth for deep MRF image reconstruction. NLEI-MRF extends the recent Equivariant Imaging framework to the MRF nonlinear inverse problem. Only compressed-sampled MRF scans are used for training. NLEI-MRF learns tissue mapping using spatiotemporal priors: spatial priors are obtained from the invariance of MRF data to a group of geometric image transformations, while temporal priors are obtained from a nonlinear Bloch response model approximated by a pre-trained neural network. Tested retrospectively on two acquisition settings, we observe that NLEI-MRF closely approaches the performance of supervised learning.

AB - Current state-of-the-art reconstruction for quantitative tissue maps from fast, compressive, Magnetic Resonance Fingerprinting (MRF), use supervised deep learning, with the drawback of requiring high-fidelity ground truth tissue map training data which is limited. This paper proposes NonLinear Equivariant Imaging for MRF (NLEIMRF), a self-supervised learning approach to eliminate the need for ground truth for deep MRF image reconstruction. NLEI-MRF extends the recent Equivariant Imaging framework to the MRF nonlinear inverse problem. Only compressed-sampled MRF scans are used for training. NLEI-MRF learns tissue mapping using spatiotemporal priors: spatial priors are obtained from the invariance of MRF data to a group of geometric image transformations, while temporal priors are obtained from a nonlinear Bloch response model approximated by a pre-trained neural network. Tested retrospectively on two acquisition settings, we observe that NLEI-MRF closely approaches the performance of supervised learning.

KW - Compressed Sensing

KW - Equivariant Imaging

KW - Inverse Problems

KW - Magnetic Resonance Fingerprinting

KW - Quantitative MRI

KW - Self-Supervised Deep Learning

UR - https://www.scopus.com/pages/publications/85172115918

U2 - 10.1109/ISBI53787.2023.10230440

DO - 10.1109/ISBI53787.2023.10230440

M3 - Chapter in a published conference proceeding

AN - SCOPUS:85172115918

T3 - Proceedings - International Symposium on Biomedical Imaging

BT - 2023 IEEE International Symposium on Biomedical Imaging, ISBI 2023

PB - IEEE

CY - U. S. A.

Y2 - 18 April 2023 through 21 April 2023

ER -

Nonlinear Equivariant Imaging: Learning Multi-Parametric Tissue Mapping without Ground Truth for Compressive Quantitative MRI

Abstract

Publication series

Conference

Bibliographical note

Funding

Keywords

ASJC Scopus subject areas

Access to Document

Other files and links

Fingerprint

Cite this