Paper Info Reviews Meta-Review Author Feedback Post-rebuttal Meta-Reviews

Authors

Khoi Minh Huynh, Wei-Tang Chang, Sang Hun Chung, Yong Chen, Yueh Lee, Pew-Thian Yap

Abstract

In magnetic resonance imaging (MRI), noise is a limiting factor for higher spatial resolution and a major cause of prolonged scan time, owing to the need for repeated scans. Improving the signal-to- noise ratio is therefore key to faster and higher-resolution MRI. Here we propose a method for mapping and reducing noise in MRI by leveraging the inherent redundancy in complex-valued multi-channel MRI data. Our method leverages a provably optimal strategy for shrinking the singular values of a data matrix, allowing it to outperform state-of-the-art methods such as Marchenko-Pastur PCA in noise reduction. Our method reduces the noise floor in brain diffusion MRI by 5-fold and fast spiral lung 19F MRI by X-fold. Our framework is fast and does not require training and hyper-parameter tuning, therefore providing a convenient means for improving SNR in MRI.

Link to paper

DOI: https://doi.org/10.1007/978-3-030-87231-1_19

SharedIt: https://rdcu.be/cyhUK

Link to the code repository

N/A

Link to the dataset(s)

N/A

Reviews

Review #1

Please describe the contribution of the paper

The paper is focused on reducing noise in MRI by utilizing the inherent redundancy in complex-valued multi-channel MRI data.
Please list the main strengths of the paper; you should write about a novel formulation, an original way to use data, demonstration of clinical feasibility, a novel application, a particularly strong evaluation, or anything else that is a strong aspect of this work. Please provide details, for instance, if a method is novel, explain what aspect is novel and why this is interesting.
- Novel formulation to reduce the noise in DW-MRI data
- demonstration on in-vivo brain and lung data. Also synthetic data
- Results are promising as they reduce the noise by 5-fold
Please list the main weaknesses of the paper. Please provide details, for instance, if you think a method is not novel, explain why and provide a reference to prior work.
- No major weakness seen with the work
Please rate the clarity and organization of this paper

Good
Please comment on the reproducibility of the paper. Note, that authors have filled out a reproducibility checklist upon submission. Please be aware that authors are not required to meet all criteria on the checklist - for instance, providing code and data is a plus, but not a requirement for acceptance

The experiments have not been repeated.
Please provide detailed and constructive comments for the authors. Please also refer to our Reviewer’s guide on what makes a good review: https://miccai2021.org/en/REVIEWER-GUIDELINES.html
- Testing on additional public datasets could show the viability of the method even better
- Fig 4. and 5. please revise the term ‘noisy’ or rephrase. The whole sentence does not make sense “Noisy and denoising results given by different methods”
- Very minor typos
Please state your overall opinion of the paper

Probably accept (7)
Please justify your recommendation. What were the major factors that led you to your overall score for this paper?

Good formulation, empirical evidence by demonstration on in-vivo data that the method works on brain and lung MRI data.
What is the ranking of this paper in your review stack?

2
Number of papers in your stack

5
Reviewer confidence

Somewhat confident

Review #2

Please describe the contribution of the paper

The paper describes a denoising method for MRI data that leverages the signal redundancy across image volumes and across coil channels. To this end, the work uses a combination of phase correction, noise level estimation and singular value shrinkage. The method scores best in a comparison against against 3 other related approaches.
Please list the main strengths of the paper; you should write about a novel formulation, an original way to use data, demonstration of clinical feasibility, a novel application, a particularly strong evaluation, or anything else that is a strong aspect of this work. Please provide details, for instance, if a method is novel, explain what aspect is novel and why this is interesting.
- A complete approach that combines the best ideas of several prior works.
- Thorough evaluation in simulated and in vivo data.
Please list the main weaknesses of the paper. Please provide details, for instance, if you think a method is not novel, explain why and provide a reference to prior work.
- Two highly relevant related works are not referenced or discussed. These are: 1) Cordero-Grande et al. (NeuroImage, 200:391-404) which used denoising in complex dMRI data with phase demodulation and optimal shrinkage, and 2) Lemberskiy et al. (Proc. ISMRM 2019) which also applied MP-PCA denoising to the coil channel images during reconstruction. The proposed work is the direct combination of both works, so it seems fair to cite them and acknowledge their contributions.
- In light of the previous point, the validation would be more informative if these prior works were included in the comparison. The “in house” method MCC-MP seems to mimic Lemberskiy et al., but a comparison with applying optimal shrinkage denoising to the complex phase-corrected data after reconstruction (as in Cordero-Grande et al.) is missing and would offer great insight into the added value of including the raw coil channels.
Please rate the clarity and organization of this paper

Very Good
Please comment on the reproducibility of the paper. Note, that authors have filled out a reproducibility checklist upon submission. Please be aware that authors are not required to meet all criteria on the checklist - for instance, providing code and data is a plus, but not a requirement for acceptance
- It is not clear from the author statement if the code will be made available or not.
Please provide detailed and constructive comments for the authors. Please also refer to our Reviewer’s guide on what makes a good review: https://miccai2021.org/en/REVIEWER-GUIDELINES.html

1) Expanding slightly on what was said above, the experimental comparison would be strengthened if you could also compare to a method that first reconstructs the MRI phase and magnitude images and then applies denoising of these complex data with phase correction and optimal singular value shrinkage. This would not only provide a comparison to published work, but would also offer insight into the added value of including the individual coil channels. Could you discuss the additional information provided in the individual coil channels, given that parallel MR reconstruction already enforces a (denoising) rank reduction? 2) Please avoid the notation “B0s” when referring to non-diffusion weighted images (b = 0), to avoid confusion with the B0-field in MRI.
Please state your overall opinion of the paper

borderline reject (5)
Please justify your recommendation. What were the major factors that led you to your overall score for this paper?

This is interesting and well described work. To reach an “accept” recommendation, the authors need to put their contributions into the relevant context of prior works.
What is the ranking of this paper in your review stack?

2
Number of papers in your stack

3
Reviewer confidence

Very confident

Review #3

Please describe the contribution of the paper

The authors propose a method for denoising multi-channel MRI. The two advances are use of the multi-channel (rather than channel combined) data, and use of an optimal stratgey for shrinking the singular values of a data matrix. Results in simulation and real data of the head and lungs show substantial improvements in noise-reduction compared to existing methods, especially for cases when the number of volumes acquired is small.
Please list the main strengths of the paper; you should write about a novel formulation, an original way to use data, demonstration of clinical feasibility, a novel application, a particularly strong evaluation, or anything else that is a strong aspect of this work. Please provide details, for instance, if a method is novel, explain what aspect is novel and why this is interesting.
- Very clearly written
- Novel methodology
- Well validated
Please list the main weaknesses of the paper. Please provide details, for instance, if you think a method is not novel, explain why and provide a reference to prior work.

-
Please rate the clarity and organization of this paper

Excellent
Please comment on the reproducibility of the paper. Note, that authors have filled out a reproducibility checklist upon submission. Please be aware that authors are not required to meet all criteria on the checklist - for instance, providing code and data is a plus, but not a requirement for acceptance

The statement is filled out correctly. It is a shame that code is not made available.
Please provide detailed and constructive comments for the authors. Please also refer to our Reviewer’s guide on what makes a good review: https://miccai2021.org/en/REVIEWER-GUIDELINES.html
All my comments are fairly minor:
- I would have appreciated some results in simulation showing how the noise-removal methods affected any DW-MRI metrics (e.g. FA, MD) fit to the data. Is it possible these methods can bias downstream metrics?
- Abstract - I think there is a number missing (text says X-fold)
- Hard to make out red arrows in Fig 1. Scale bars would also be helpful for the phase and noise maps.
- Is the choice of weights in Eq 11 to reduce Gibbs aretefacts already present in the reconstructed data, or to prevent the introduction of new Gibbs artefacts by the noise removal process?
- Conclusions: Tunning -> Tuning
Please state your overall opinion of the paper

strong accept (9)
Please justify your recommendation. What were the major factors that led you to your overall score for this paper?

Novel work with strong results.
What is the ranking of this paper in your review stack?

1
Number of papers in your stack

5
Reviewer confidence

Confident but not absolutely certain

Primary Meta-Review

Please provide your assessment of this work, taking into account all reviews. Summarize the key strengths and weaknesses of the paper and justify your recommendation. In case you deviate from the reviewers’ recommendations, explain in detail the reasons why. In case of an invitation for rebuttal, clarify which points are important to address in the rebuttal.
This paper Nr 633 presents a method to reduce the noise in MRI relying on utilizing inherent redundancy in compex-values multi-channel MRI data. The problem addresses a highly relevant problem, the trade-off between spatial resolution and noise. Major strengths of the paper as pointed out by the reviewers are:
- The paper presents a method which can be applied for many applications, two very different ones are already shown and evaluated in the paper itself (brain & lung MRI with different methods applied.)
- The thorough evaluation and strong results are pointed out by all reviewers, sicj as the achieved 5-fold acceleration and thorough evaluation in simulated and in-vivo MRI data of two different applications
- The reviewers point out that the paper is well written. The Figures are well suited to illustrate the points outlined, it is made easy to read and understand for the reader. Some minor flaws with the Figures are pointed out in the minor comments.
The major weakness pointed out by the reviewers was that prior work is not cited appropriately, eg the papers by Cordero-Grande or Lemberskiy. This is however highly relevant as the method is an elegant combination of elements proposed before.

Minor comments:
- Please also refer to the minor comments in the three reviews.
- This paper needs a thorough re-read! There are typos, unfinished sentences and some other interesting elements (eg in the abstract X-fold reduction or “Noisy and denoising” (instead of denoised) in the captions of Fig 4 and 5.).
- The Figures need updating (see comments of reviewer 4, eg related to the arrows in fig.1 etc)
What is the ranking of this paper in your stack? Use a number between 1 (best paper in your stack) and n (worst paper in your stack of n papers).

3

Author Feedback

We thank the reviewers and the meta-reviewer for the valuable comments. We would like to address all the major comments below. Minor suggestions for improving clarity will be incorporated in the camera-ready version of the paper.

Citing Lemberskiy et al. (Proc. ISMRM 2019) The in-house MCC-MP approach used to compare with our proposed method is inspired by the work of Lemberskiy et al. on using MP-PCA on multi-channel data. We will cite this abstract in the final version of our paper.

Citing Cordero-Grande et al. The work of Cordero-Grande et al., although uses the same term ‘optimal shrinkage’, is actually based on the spike covariance model (manipulating the eigenvalue of the covariance of the signal matrix). In contrast, our method manipulates the singular value of the signal matrix itself. Our method has different assumptions and is not a derivation of the work of Cordero-Grande et al. Our method also differs from Cordero-Grande et al.’s method in terms of pre-processing steps. We will cite Cordero-Grande et al. and discuss the differences.

Denoising of phase-corrected data Our method applies shrinkage on data after removing the background phase.

back to top

Noise Mapping and Removal in Complex-Valued Multi-Channel MRI via Optimal Shrinkage of Singular Values