Authors

Feng Liu, Guihong Wan, Yevgeniy R. Semenov, Patrick L. Purdon

Abstract

Electrophysiological Source Imaging (ESI) refers to the process of localizing the brain source activation patterns given measured Electroencephalography (EEG) or Magnetoencephalography (MEG) signal from the scalp. Recent studies have focused on designing sophisticated neurophysiologically plausible regularizations or efficient estimation frameworks to solve the ESI problem, with the underlying assumption that brain source activation has some specific structures. Estimation of both source location and its extents is important in clinical applications. However, estimating the high dimensional extended location is challenging due to the highly coherent columns in the leadfield matrix, resulting in a reconstructed spiky spurious sources. In this work, we describe an efficient and accurate framework by exploiting the graph structure defined in the 3D mesh of the brain. Specifically, we decompose the graph signal representation in the source space into low-, medium-, and high-frequency subspaces, and project the source signal into the graph low-frequency subspace. We further introduce a low-rank representation with temporal graph regularization in the projected space to build the ESI framework, which can be efficiently solved. Experiments with simulated data and real world EEG data demonstrated the superiority of the proposed paradigm for estimating brain source extents.

Link to paper

DOI: https://link.springer.com/chapter/10.1007/978-3-031-16431-6_10

SharedIt: https://rdcu.be/cVD4R

Link to the code repository

N/A

Link to the dataset(s)

N/A

Reviews

Review #1

Please describe the contribution of the paper

This paper addresses the ill-posed problem of source reconstruction in EEG or MEG. The idea of the method is to be less sensitive to spatial high frequency activation. The method proposes a law rank representation to estimate the different parameters on a projected subspace spanned by a low-frequency graph basis.
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.

The strength of this paper is to project the problem in a subspace with lower frequency, and ask for temporal and spatial regularisation, that is feasible thanks to the representation of the problem with graphs.
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.

There is too much supplementary material, better write a journal paper to be more comfortable, this would allow you to give more details in the method, or limit your paper to synthetic data analysis. As it is, it is frustrating to have only partial information.
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

No code, no data available, nor algorithm given in the paper. However, the paper gives enough details to re-implement the method. The parameter tuning is not mentioned, that provide from reproducing results.
Please provide detailed and constructive comments for the authors. Please also refer to our Reviewer’s guide on what makes a good review: https://conferences.miccai.org/2022/en/REVIEWER-GUIDELINES.html
- Did you try other metrics to evaluate the performance of the method? To me, when looking at Fig2 (3rd row), the MNE method gives a more suitable result than the proposed method that eventually covers a quarter of the hemisphere.
- I do not understand why on real data, the proposed method gives such a sparse solution compared to other. This is counter-intuitive when regarding at the results on synthetic data.
- On real data you mention the highly diffuse activation of other method, but on synthetic data your method is the most diffuse.
- There is too much supplementary material, better write a journal paper to be more comfortable, this would allow you to give more details in the method, or limit your paper to synthetic data analysis. As it is, it is frustrating to have only partial information.
Rate the paper on a scale of 1-8, 8 being the strongest (8-5: accept; 4-1: reject). Spreading the score helps create a distribution for decision-making

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

see comments above
Number of papers in your stack

1
What is the ranking of this paper in your review stack?

1
Reviewer confidence

Confident but not absolutely certain
[Post rebuttal] After reading the author’s rebuttal, state your overall opinion of the paper if it has been changed

N/A
[Post rebuttal] Please justify your decision

N/A

Review #2

Please describe the contribution of the paper

The authors provide a novel method for estimation of both source location and extents . They provide a new that exploits the graph structure defined in the 3D mesh of the brain by separating the graph signal into different frequency subspaces, where they project the signal.
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.

The method proposed is novel and the results presented are very convincing for the need of the method. The rational - starting point of the method used makes absolute sense and overall is a well written- well structured paper.
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.

The main weakness of the paper is the lack of information about the tuning of the parameters and their optimal selection. In methods were multiple regularizers are employed a critical point is the difficulty of the selection of the regularizers parameters (a,b,g etc). Especially when comparison among such methods is made it needs to be sure that all the parameters are optimized (for fairness).
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

The authors have done a very nice work in terms of reproducibility and the supplementary text helps a lot towards this direction, especially if the code will be released.

One minor note, the authors answered positive in the question:

The average runtime for each result, or estimated energy cost.

But the average runtime is not mentioned
Please provide detailed and constructive comments for the authors. Please also refer to our Reviewer’s guide on what makes a good review: https://conferences.miccai.org/2022/en/REVIEWER-GUIDELINES.html
Will start for more details about the main weakness of the paper. The proposed method includes a,b, and λ . Τhe authors they do not mention a lot how the tuning of this parameters affects the results and how susceptible to error is based on a suboptimal selection of the parameters. Furthermore when coming to comparison between other methods with less parameters (e.g. I think sloreta has only 1 regularizer) is even more important to include the influence and easiness of tuning them.

Other comments of less importance
- Fig. S.5.1 I n the case of 10 db. The performance difference between the proposed method and all the others seem significant while from table 1 this is not the case (From auc the method is not even optimal). Could you elaborate on this, might be an indication that the metrics are not optimal and some metric of mutual information - or correlation might be better?
- “We set the length of EEG to 1 second” - Not very clear to me you mean the whole EEG was 1 second? Why such a small choice is it much more difficult the the graph setting will be used in bigger segments?
- In case where you have two distinct areas of activation (simultaneous) e.g. left and right temporal lobes, instead of 1. How would this affect the performance of the method? Especially the “Forcing” of neighboring signal (which helps in the case of single area of activation, as well as the low rankness when you have multiple “areas of activation” (e.g. temporoparietal network)
- Define tr (is trace ok but you need to define it)
Rate the paper on a scale of 1-8, 8 being the strongest (8-5: accept; 4-1: reject). Spreading the score helps create a distribution for decision-making

6
Please justify your recommendation. What were the major factors that led you to your overall score for this paper?

The paper is of good quality and of interest for publication the idea behind the use of the method has ground and the results presented are very good since outperform methods that are being used for year as sota. There is a need for clarifications on the comparison between those methods especially for the tuning of the hyperparameters.
Number of papers in your stack

2
What is the ranking of this paper in your review stack?

1
Reviewer confidence

Somewhat Confident
[Post rebuttal] After reading the author’s rebuttal, state your overall opinion of the paper if it has been changed

N/A
[Post rebuttal] Please justify your decision

N/A

Review #3

Please describe the contribution of the paper

This paper investigates extended electrophysiological source imaging with spatial graph filters. The simulation tests have been carried out in detail.
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.

In general, this paper is technically sound and the topic is interesting.
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.

Here, there are some comments of this reviewer: 1 In the introduction section, the literature review must be strengthened. Avoid lumping references as in [2,21,24,32], [30, 22, 5, 4, 1] and all others. Instead summarize the main contribution of each referenced paper in a separate sentence 2 In this work, how to guarantee of the convergence of the ADMM used in the final algorithm? 3 The proposed method might be sensitive to the values of its main controlling parameters. How did you determine these parameters? Please elaborate on that. 4 In practical applications, noises may be non-Gaussian noises. Have you considered such non-Gaussian noises? Please discuss how this would impact the results and conclusions of this study.
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

I think the reproducibility of the paper is good.
Please provide detailed and constructive comments for the authors. Please also refer to our Reviewer’s guide on what makes a good review: https://conferences.miccai.org/2022/en/REVIEWER-GUIDELINES.html

Please specify details of the computing platform, programming language and parameter settings used in this study.
Rate the paper on a scale of 1-8, 8 being the strongest (8-5: accept; 4-1: reject). Spreading the score helps create a distribution for decision-making

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

In general, this paper is well written. The major factors are as follows: 1 The contributions are clearly demonstrated. 2 The paper is technially sound as the verifications of this work is sufficient. 3 The presentation is acceptable.
Number of papers in your stack

4
What is the ranking of this paper in your review stack?

1
Reviewer confidence

Confident but not absolutely certain
[Post rebuttal] After reading the author’s rebuttal, state your overall opinion of the paper if it has been changed

N/A
[Post rebuttal] Please justify your decision

N/A

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.

The overall recommendation of this paper towards accept. All reviewers positively commented on the novelty and technical soundness of the proposed method. The authors are strongly encouraged to incorporate all reviewers’ constructive feedback carefully into a revised version.
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). If this paper is among the bottom 30% of your stack, feel free to use NR (not ranked).

2

Author Feedback

N/A

back to top

Extended Electrophysiological Source Imaging with Spatial Graph Filters