MRzero - Automated discovery of MRI sequences using supervised learning
© 2021 The Authors. Magnetic Resonance in Medicine published by Wiley Periodicals LLC on behalf of International Society for Magnetic Resonance in Medicine..
PURPOSE: 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.
METHODS: 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, T1 and T2 , and ΔB0 . As a proof of concept, we use both conventional MR images and T1 maps as targets and optimize from scratch using the loss defined by data fidelity, SAR penalty, and scan time.
RESULTS: 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.
CONCLUSIONS: Automated MR sequence generation is possible based on differentiable Bloch equation simulations and a supervised learning approach.
| Medienart: |
E-Artikel |
|---|
| Erscheinungsjahr: |
2021 |
|---|---|
| Erschienen: |
2021 |
| Enthalten in: |
Zur Gesamtaufnahme - volume:86 |
|---|---|
| Enthalten in: |
Magnetic resonance in medicine - 86(2021), 2 vom: 01. Aug., Seite 709-724 |
| Sprache: |
Englisch |
|---|
| Beteiligte Personen: |
Loktyushin, A [VerfasserIn] |
|---|
| Links: |
|---|
| Themen: |
AUTOSEQ |
|---|
| Anmerkungen: |
Date Completed 20.05.2021 Date Revised 20.05.2021 published: Print-Electronic Citation Status MEDLINE |
|---|
| doi: |
10.1002/mrm.28727 |
|---|
| funding: |
|
|---|---|
| Förderinstitution / Projekttitel: |
|
| PPN (Katalog-ID): |
NLM323115349 |
|---|
| LEADER | 01000naa a22002652 4500 | ||
|---|---|---|---|
| 001 | NLM323115349 | ||
| 003 | DE-627 | ||
| 005 | 20231225183501.0 | ||
| 007 | cr uuu---uuuuu | ||
| 008 | 231225s2021 xx |||||o 00| ||eng c | ||
| 024 | 7 | |a 10.1002/mrm.28727 |2 doi | |
| 028 | 5 | 2 | |a pubmed24n1077.xml |
| 035 | |a (DE-627)NLM323115349 | ||
| 035 | |a (NLM)33755247 | ||
| 040 | |a DE-627 |b ger |c DE-627 |e rakwb | ||
| 041 | |a eng | ||
| 100 | 1 | |a Loktyushin, A |e verfasserin |4 aut | |
| 245 | 1 | 0 | |a MRzero - Automated discovery of MRI sequences using supervised learning |
| 264 | 1 | |c 2021 | |
| 336 | |a Text |b txt |2 rdacontent | ||
| 337 | |a ƒaComputermedien |b c |2 rdamedia | ||
| 338 | |a ƒa Online-Ressource |b cr |2 rdacarrier | ||
| 500 | |a Date Completed 20.05.2021 | ||
| 500 | |a Date Revised 20.05.2021 | ||
| 500 | |a published: Print-Electronic | ||
| 500 | |a Citation Status MEDLINE | ||
| 520 | |a © 2021 The Authors. Magnetic Resonance in Medicine published by Wiley Periodicals LLC on behalf of International Society for Magnetic Resonance in Medicine. | ||
| 520 | |a PURPOSE: 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 | ||
| 520 | |a METHODS: 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, T1 and T2 , and ΔB0 . As a proof of concept, we use both conventional MR images and T1 maps as targets and optimize from scratch using the loss defined by data fidelity, SAR penalty, and scan time | ||
| 520 | |a RESULTS: 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 | ||
| 520 | |a CONCLUSIONS: Automated MR sequence generation is possible based on differentiable Bloch equation simulations and a supervised learning approach | ||
| 650 | 4 | |a Journal Article | |
| 650 | 4 | |a Research Support, Non-U.S. Gov't | |
| 650 | 4 | |a AUTOSEQ | |
| 650 | 4 | |a MR simulation | |
| 650 | 4 | |a automatic MR | |
| 650 | 4 | |a differentiable Bloch equation | |
| 650 | 4 | |a machine learning | |
| 700 | 1 | |a Herz, K |e verfasserin |4 aut | |
| 700 | 1 | |a Dang, N |e verfasserin |4 aut | |
| 700 | 1 | |a Glang, F |e verfasserin |4 aut | |
| 700 | 1 | |a Deshmane, A |e verfasserin |4 aut | |
| 700 | 1 | |a Weinmüller, S |e verfasserin |4 aut | |
| 700 | 1 | |a Doerfler, A |e verfasserin |4 aut | |
| 700 | 1 | |a Schölkopf, B |e verfasserin |4 aut | |
| 700 | 1 | |a Scheffler, K |e verfasserin |4 aut | |
| 700 | 1 | |a Zaiss, M |e verfasserin |4 aut | |
| 773 | 0 | 8 | |i Enthalten in |t Magnetic resonance in medicine |d 1993 |g 86(2021), 2 vom: 01. Aug., Seite 709-724 |w (DE-627)NLM01259413X |x 1522-2594 |7 nnns |
| 773 | 1 | 8 | |g volume:86 |g year:2021 |g number:2 |g day:01 |g month:08 |g pages:709-724 |
| 856 | 4 | 0 | |u http://dx.doi.org/10.1002/mrm.28727 |3 Volltext |
| 912 | |a GBV_USEFLAG_A | ||
| 912 | |a GBV_NLM | ||
| 951 | |a AR | ||
| 952 | |d 86 |j 2021 |e 2 |b 01 |c 08 |h 709-724 | ||