MRzero - Automated discovery of MRI sequences using supervised learning.
AUTOSEQ
MR simulation
automatic MR
differentiable Bloch equation
machine learning
Journal
Magnetic resonance in medicine
ISSN: 1522-2594
Titre abrégé: Magn Reson Med
Pays: United States
ID NLM: 8505245
Informations de publication
Date de publication:
08 2021
08 2021
Historique:
revised:
15
01
2021
received:
04
02
2020
accepted:
20
01
2021
pubmed:
24
3
2021
medline:
21
5
2021
entrez:
23
3
2021
Statut:
ppublish
Résumé
A supervised learning framework is proposed to automatically generate MR sequences and corresponding reconstruction based on the target contrast of interest. Combined with a flexible, task-driven cost function this allows for an efficient exploration of novel MR sequence strategies. The scanning and reconstruction process is simulated end-to-end in terms of RF events, gradient moment events in x and y, and delay times, acting on the input model spin system given in terms of proton density, In a first attempt, MRzero learns gradient and RF events from zero, and is able to generate a target image produced by a conventional gradient echo sequence. Using a neural network within the reconstruction module allows arbitrary targets to be learned successfully. Experiments could be translated to image acquisition at the real system (3T Siemens, PRISMA) and could be verified in the measurements of phantoms and a human brain in vivo. Automated MR sequence generation is possible based on differentiable Bloch equation simulations and a supervised learning approach.
Types de publication
Journal Article
Research Support, Non-U.S. Gov't
Langues
eng
Sous-ensembles de citation
IM
Pagination
709-724Informations de copyright
© 2021 The Authors. Magnetic Resonance in Medicine published by Wiley Periodicals LLC on behalf of International Society for Magnetic Resonance in Medicine.
Références
CarrHY. Steady-state free precession in nuclear magnetic resonance. Phys Rev. 1958;112:1693-1701.
HaaseA, FrahmJ, MatthaeiD, HanickeW, MerboldtKD. FLASH imaging. Rapid NMR imaging using low flip-angle pulses. J Magn Reson (1969). 1986;67:258-266.
HennigJ, NauerthA, FriedburgH. RARE imaging: a fast imaging method for clinical MR. Magn Reson Med. 1986;3:823-833.
HargreavesBA, NishimuraDG, ConollySM. Time-optimal multidimensional gradient waveform design for rapid imaging. Magn Reson Med. 2004;51:81-92.
LustigM, KimSJ, PaulyJM. A fast method for designing time-optimal gradient waveforms for arbitrary k-space trajectories. IEEE Trans Med Imaging. 2008;27:866-873.
RouxPL, HinksRS. Stabilization of echo amplitudes in FSE sequences. Magn Reson Med. 1993;30:183-190.
ShinnarM, EleffS, SubramanianH, LeighJS. The synthesis of pulse sequences yielding arbitrary magnetization vectors. Mag Reson Med. 1989;12:74-80.
BahadirCD, DalcaAV, SabuncuMR. Learning-Based Optimization of the Under-Sampling Pattern in MRI. arXiv 2019;arXiv eprint: 1901.01960.
JinKH, UnserM, YiKM. Self-Supervised Deep Active Accelerated MRI.
SherryF, BenningM, los ReyesJCD, et al. Learning the Sampling Pattern for MRI. arXiv 2019;arXiv eprint: 1906.08754.
WeissT, SenoufO, VedulaS, MichailovichO, ZibulevskyM, BronsteinA. PILOT: Physics-Informed Learned Optimal Trajectories for Accelerated MRI. arXiv 2019;arXiv eprint: 1909.05773.
WeissT, VedulaS, SenoufO, BronsteinA, MichailovichO, ZibulevskyM. Learning Fast Magnetic Resonance Imaging. arXiv 2019;arXiv eprint: 1905.09324.
ShinF. Deep Reinforcement Learning Designed RF Pulse. 2019. https://index.mirasmart.com/ISMRM2019/PDFfiles/0757.html
Walker-Samuel. Using Deep Reinforcement Learning to Actively, Adaptively and Autonomously Control a Simulated MRI Scanner. 2019. https://cds.ismrm.org/protected/19MPresentations/abstracts/0473.html.
ZhuB, LiuJ, KoonjooN, RosenB, RosenM. AUTOmated pulse SEQuence generation (AUTOSEQ) using Bayesian reinforcement learning in an MRI physics simulation environment. Proceedings 26th Annual Meeting ISMRM, Paris, France; 2018:16-21.
ZhuB, LiuJ, KoonjooN, RosenB, RosenMS. AUTOmated pulse SEQuence generation (AUTOSEQ) and neural network decoding for fast quantitative MR parameter measurement using continuous and simultaneous RF transmit and receive. ISMRM Annual Meeting & Exhibition. Vol 1090. Montréal, QC, Canada. 2019.
PaszkeA, GrossS, ChintalaS, et al. Automatic Differentiation in PyTorch. 2017.
AbadiM, AgarwalA, BarhamP, et al. TensorFlow: Large-Scale Machine Learning on Heterogeneous Systems. 2015.
LaytonKJ, KrobothS, JiaF, et al. Pulseq: a rapid and hardware-independent pulse sequence prototyping framework. Magn Reson Med. 2017;77:1544-1552. https://onlinelibrary.wiley.com/doi/abs/10.1002/mrm.26235.
KingmaDP, BaJ. Adam: A Method for Stochastic Optimization. arXiv 2014;arXiv eprint: 1412.6980.
CocoscoCA, KollokianV, KwanRKS, EvansAC. BrainWeb: Online Interface to a 3D MRI Simulated Brain Database. NeuroImage. 1997;5(4), part 2/4:S425.
BusseRF, HariharanH, VuA, BrittainJH. Fast spin echo sequences with very long echo trains: design of variable refocusing flip angle schedules and generation of clinical T2 contrast. Magn Reson Med. 2006;55:1030-1037.
HennigJSK, WeigelM. Multiecho sequences with variable refocusing flip angles: optimization of signal behavior using smooth transitions between pseudo steady states (TRAPS). Magn Reson Med. 2003;49:527-535.
SbrizziA, HoogduinH, HajnalJV, van denBergCA, LuijtenPR, MalikSJ. Optimal control design of turbo spin-echo sequences with applications to parallel-transmit systems. Mag Reson Med. 2017;77:361-373.
PruessmannKP, WeigerM, BörnertP, BoesigerP. Advances in sensitivity encoding with arbitrary k-space trajectories. Magn Reson Med. 2001;46:638-651.
SeegerM, NickischH, PohmannR, SchölkopfB. Optimization of k-space trajectories for compressed sensing by Bayesian experimental design. Magn Reson Med. 2010;63:116-126.
KnollF, HammernikK, ZhangC, et al. Deep Learning Methods for Parallel Magnetic Resonance Image Reconstruction. arXiv 2019;arXiv eprint: 1904.01112.
MaierAK, SybenC, StimpelB, WürflT, HoffmannM, SchebeschF, et al. Learning with known operators reduces maximum error bounds. Nat Mach Intelligence. 2019;1:373-380.
ZhuB, LiuJZ, CauleySF, RosenBR, RosenMS. Image reconstruction by domain-transform manifold learning. Nature. 2018;555:487.
ConollyS, NishimuraD, MacovskiA. Optimal control solutions to the magnetic resonance selective excitation problem. IEEE Trans Med Imaging. 1986;5:106-115.
LapertM, ZhangY, JanichMA, GlaserSJ, SugnyD. Exploring the physical limits of saturation contrast in magnetic resonance imaging. Sci Rep. 2012;2:1-5.
RundA, AignerCS, KunischK, StollbergerR. Magnetic resonance RF pulse design by optimal control with physical constraints. IEEE Trans Med Imaging. 2017;37:461-472.
VindingMS, MaximovII, TošnerZ, NielsenNC. Fast numerical design of spatial-selective rf pulses in MRI using Krotov and quasi-Newton based optimal control methods. J Chem Phys. 2012;137:054203.
CohenO, RosenMS. Algorithm comparison for schedule optimization in MR fingerprinting. Magn Reson Imaging. 2017;41:15-21.
KörzdörferG, KirschR, LiuK, et al. Reproducibility and repeatability of MR fingerprinting relaxometry in the human brain. Radiology. 2019;292:429-437.
MaD, GulaniV, SeiberlichN, et al. Magnetic resonance fingerprinting. Nature. 2013;495:187. https://doi.org/10.1038/nature11971.
McGivneyDF, BoyaciogluR, JiangY, et al. Magnetic resonance fingerprinting review part 2: technique and directions. J Magn Reson Imaging. 2019. https://onlinelibrary.wiley.com/doi/abs/10.1002/jmri.26877.
PoormanME, MartinMN, MaD, et al. Magnetic resonance fingerprinting Part 1: (potential) uses, current challenges, and recommendations. J Magn Reson Imaging. 2019. https://onlinelibrary.wiley.com/doi/abs/10.1002/jmri.26836.
VirtueP, StellaXY, LustigM. Better than real: complex-valued neural nets for MRI fingerprinting. 2017 IEEE International Conference on Image Processing (ICIP) IEEE. 2017:3953-3957.
TeixeiraRPA, MalikSJ, HajnalJV. Joint system relaxometry (JSR) and Cramer-Rao lower bound optimization of sequence parameters: a framework for enhanced precision of DESPOT T1 and T2 estimation. Mag Reson Med. 2018;79:234-245.
RaviKS, PotdarS, PoojarP, et al. Pulseq-graphical programming interface: open source visual environment for prototyping pulse sequences and integrated magnetic resonance imaging algorithm development. Magn Reson Imaging. 2018;52:9-15.