Literature DB >> 35344614

Cramér-Rao bound-informed training of neural networks for quantitative MRI.

Xiaoxia Zhang1,2, Quentin Duchemin3, Kangning Liu4, Cem Gultekin5, Sebastian Flassbeck1,2, Carlos Fernandez-Granda4,5, Jakob Assländer1,2.   

Abstract

PURPOSE: To improve the performance of neural networks for parameter estimation in quantitative MRI, in particular when the noise propagation varies throughout the space of biophysical parameters. THEORY AND METHODS: A theoretically well-founded loss function is proposed that normalizes the squared error of each estimate with respective Cramér-Rao bound (CRB)-a theoretical lower bound for the variance of an unbiased estimator. This avoids a dominance of hard-to-estimate parameters and areas in parameter space, which are often of little interest. The normalization with corresponding CRB balances the large errors of fundamentally more noisy estimates and the small errors of fundamentally less noisy estimates, allowing the network to better learn to estimate the latter. Further, proposed loss function provides an absolute evaluation metric for performance: A network has an average loss of 1 if it is a maximally efficient unbiased estimator, which can be considered the ideal performance. The performance gain with proposed loss function is demonstrated at the example of an eight-parameter magnetization transfer model that is fitted to phantom and in vivo data.
RESULTS: Networks trained with proposed loss function perform close to optimal, that is, their loss converges to approximately 1, and their performance is superior to networks trained with the standard mean-squared error (MSE). The proposed loss function reduces the bias of the estimates compared to the MSE loss, and improves the match of the noise variance to the CRB. This performance gain translates to in vivo maps that align better with the literature.
CONCLUSION: Normalizing the squared error with the CRB during the training of neural networks improves their performance in estimating biophysical parameters.
© 2022 International Society for Magnetic Resonance in Medicine.

Entities:  

Keywords:  deep learning; magnetic resonance fingerprinting; magnetic transfer; objective function; parameter estimation; quantitative MRI

Mesh:

Year:  2022        PMID: 35344614      PMCID: PMC9050814          DOI: 10.1002/mrm.29206

Source DB:  PubMed          Journal:  Magn Reson Med        ISSN: 0740-3194            Impact factor:   3.737


  31 in total

Review 1.  The clinico-radiological paradox in multiple sclerosis revisited.

Authors:  Frederik Barkhof
Journal:  Curr Opin Neurol       Date:  2002-06       Impact factor: 5.710

2.  Volume and iron content in basal ganglia and thalamus.

Authors:  Patrice Péran; Andrea Cherubini; Giacomo Luccichenti; Gisela Hagberg; Jean-François Démonet; Olivier Rascol; Pierre Celsis; Carlo Caltagirone; Gianfranco Spalletta; Umberto Sabatini
Journal:  Hum Brain Mapp       Date:  2009-08       Impact factor: 5.038

3.  MR fingerprinting Deep RecOnstruction NEtwork (DRONE).

Authors:  Ouri Cohen; Bo Zhu; Matthew S Rosen
Journal:  Magn Reson Med       Date:  2018-04-06       Impact factor: 4.668

4.  Simultaneous acquisition of spatial harmonics (SMASH): fast imaging with radiofrequency coil arrays.

Authors:  D K Sodickson; W J Manning
Journal:  Magn Reson Med       Date:  1997-10       Impact factor: 4.668

5.  Optimal experiment design for magnetic resonance fingerprinting.

Authors:  Justin P Haldar; Kawin Setsompop; Lawrence L Wald
Journal:  Conf Proc IEEE Eng Med Biol Soc       Date:  2016-08

6.  Dictionary-Free MRI PERK: Parameter Estimation via Regression with Kernels.

Authors:  Gopal Nataraj; Jon-Fredrik Nielsen; Clayton Scott; Jeffrey A Fessler
Journal:  IEEE Trans Med Imaging       Date:  2018-03-20       Impact factor: 10.048

7.  Deep Learning for Fast and Spatially Constrained Tissue Quantification From Highly Accelerated Data in Magnetic Resonance Fingerprinting.

Authors:  Zhenghan Fang; Yong Chen; Mingxia Liu; Lei Xiang; Qian Zhang; Qian Wang; Weili Lin; Dinggang Shen
Journal:  IEEE Trans Med Imaging       Date:  2019-02-13       Impact factor: 10.048

8.  Deep Learning for Magnetic Resonance Fingerprinting: A New Approach for Predicting Quantitative Parameter Values from Time Series.

Authors:  Elisabeth Hoppe; Gregor Körzdörfer; Tobias Würfl; Jens Wetzl; Felix Lugauer; Josef Pfeuffer; Andreas Maier
Journal:  Stud Health Technol Inform       Date:  2017

Review 9.  A Perspective on MR Fingerprinting.

Authors:  Jakob Assländer
Journal:  J Magn Reson Imaging       Date:  2020-04-14       Impact factor: 4.813

10.  Magnetic resonance fingerprinting.

Authors:  Dan Ma; Vikas Gulani; Nicole Seiberlich; Kecheng Liu; Jeffrey L Sunshine; Jeffrey L Duerk; Mark A Griswold
Journal:  Nature       Date:  2013-03-14       Impact factor: 49.962

View more

北京卡尤迪生物科技股份有限公司 © 2022-2023.