Authors

Xiang Gao, Xin Zhang, Lu Zhang, Xiangmin Xu, Dajiang Zhu

Abstract

Revealing structural-functional relationship is crucial issue in neuroscience study since it helps to understand brain activities. Structural Connectivity (SC) represents the fibers connection between the brain regions, which is relatively static. Functional Connectivity (FC) represents the active signal correlations between the brain regions, which is relatively dynamic and diverse. Many works predict FC from SC and achieve unique FC prediction. However, FC is diverse since it represents brain activities. In this work, we propose the MCGAN, a multi-contexts discriminator based generative adversarial network for predicting diverse FC from SC. The proposed multi-contexts discriminator provides three kinds of supervisions to strengthen the generator, i.e. edge-level, node-level and graph-level. Since FC represents the connection of the brain regions, which can be regarded as edge-based graph. We adopt edge-based graph convolution method to model the context encoding. Moreover, to introduce the diversity of generated FC, we utilize monte-carlo mean samples to bring in more FC data for training. We validate our MCGAN on Human Connextome Project (HCP) dataset and Alzheimer’s Disease Neuroimaging Initiative (ADNI) dataset. The results show that our method can generate diverse and meaningful FC from SC, revealing the one-to-many relationship between the individual SC and the multiple FC. The glamorous significance of this work is that once we have anatomical structure of brain represented by SC, we can predict diverse developments of brain activity represented by FC, which helps to reveal individual brain’s static-dynamic structural-functional mode.

Link to paper

DOI: https://doi.org/10.1007/978-3-031-43993-3_34

SharedIt: https://rdcu.be/dnwNz

Link to the code repository

N/A

Link to the dataset(s)

N/A

Reviews

Review #2

Please describe the contribution of the paper

The paper presents a unique and innovative approach to predicting functional connectivity from structural connectivity in the brain using MCGAN. The methodology and results are presented in a clear and concise manner, with detailed explanations provided by the authors. The experiments demonstrate the effectiveness of the proposed method in generating diverse and meaningful functional connectivity from structural connectivity, with potential applications in the field of neuroscience for understanding individual brain activity.
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 paper presents an innovative approach to predicting functional connectivity in the brain using MCGAN, which is well-explained and easy to understand. The methodology and results are presented clearly, and the experiments demonstrate the effectiveness of the proposed method in generating meaningful functional connectivity from structural connectivity. The potential applications of this research in the field of neuroscience make it a valuable contribution.
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.

While the paper presents an innovative approach to predicting functional connectivity from structural connectivity in the brain using MCGAN, there are some limitations that need to be addressed. The experimental results presented in Figure 4 lack detailed descriptions, and it is not clear how the authors arrived at their conclusions. Additionally, the authors could have provided more detailed discussions on the effect of the parameters, such as the number of slices, for different datasets. This would have helped to establish the robustness of the proposed method across different datasets and parameter settings. Secondly, a more detailed comparison with other state-of-the-art methods for predicting functional connectivity from structural connectivity would have helped to establish the novelty and superiority of the proposed method in a more convincing manner.
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 reproducibility of the paper allows other researchers to replicate the experiments and verify the 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/2023/en/REVIEWER-GUIDELINES.html

It would be helpful to provide more detailed explanations and justifications for the choice of hyperparameters, such as the learning rate and batch size used in the experiments. This would help other researchers to understand the decisions made and possibly fine-tune these parameters for their own experiments.

The authors could consider providing a more detailed discussion on the limitations and future directions of their proposed method. This would help readers to understand the potential drawbacks and opportunities for improvement in their approach.

Overall, the paper presents a valuable contribution to the field of neuroscience with its innovative approach and meaningful results. However, addressing the above comments could further enhance the clarity and impact of the research.
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?

Strengths of the paper: The paper presents an innovative approach to predicting functional connectivity from structural connectivity in the brain using a multi-contexts discriminator based generative adversarial network (GAN) named MCGAN. The approach is well-explained, the presentation is clear, and the experiments demonstrate the effectiveness of the proposed method in generating diverse and meaningful functional connectivity.

Weaknesses of the paper: The experimental results presented in Figure 4 lack detailed descriptions, and it is not clear how the authors arrived at their conclusions. Additionally, the authors could have provided more detailed discussions on the effect of the parameters, such as the number of slices, for different datasets.

Recommendations for improvement: The authors are recommended to provide more detailed descriptions of the experimental results in Figure 4 and discuss the effect of parameters for different datasets.
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 #4

Please describe the contribution of the paper

The authors propose a multi-context discriminator-based generative adversarial network (MCGAN) for predicting diverse functional connectivity (FC) from structural connectivity (SC). The proposed multi-context discriminator provides three kinds of supervision to strengthen the generator, namely edge-level, node-level, and graph-level. The experimental results on the Human Connectome Project (HCP) dataset and Alzheimer’s Disease Neuroimaging Initiative (ADNI) dataset demonstrate that the authors’ method can generate diverse and meaningful FC from SC, revealing the one-to-many relationship between individual SC and multiple FC.
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 authors propose a multi-context discriminator that provides edge-level, node-level, and graph-level supervision to implicitly and explicitly strengthen the generator, improving the quality and diversity of predictions.
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.

(1)The analysis of the experimental results is insufficient. Although the authors have achieved the one-to-many prediction from structural connectivity to functional connectivity, they have not revealed the structural-functional patterns between brain structural connectivity and multiple functional connectivity, and the interpretability of the model is poor. (2)The authors did not elaborate on the specific structure of the generator. The proposed MCGAN consists of a generator and a multi-contexts discriminator, but the authors did not introduce the specific structure of the generator in the method part. (3)The authors did not clarify the sample size of the dataset used nor the method of dividing the training and testing sets. (4)The authors did not explain the convergence and robustness of the model. The training process of GAN models is unstable and slow, and there may be problems with model convergence and robustness. The authors should explain how they solved these problems.
Please rate the clarity and organization of this paper

Satisfactory
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

Not very convincing. How GAN collapse?
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/2023/en/REVIEWER-GUIDELINES.html

(1)The authors provided too few evaluation metrics and no confidence interval. We suggest that the authors adopt other indicators to evaluate the diversity of the results and use methods, such as independent repetition experiments, to avoid accidental results. (2)The analysis of the experimental results is insufficient. We suggest that the authors analyze the pattern changes between the structural connectivity and multiple functional connectivity of the subjects to reveal the structural-functional mapping relationship of the brain. Moreover, the authors should analyze the differences between multiple functional connectivity of the same subject to demonstrate the diversity of the results. (3)The authors did not elaborate on the specific structure of the generator. We suggest that the authors provide more information about the generator. (4)The authors did not clarify the sample size of the dataset used nor the method of dividing the training and testing sets. We suggest that the authors provide a clearer description of the data collection, processing, and division methods. (5)The implementation details of the experiments are insufficiently described. The authors did not explain the experimental environment, hyperparameter settings (such as iteration times), and other issues. We suggest that the authors supplement the experimental details such as the experimental environment, hyperparameter settings, etc. (6)The authors did not explain the convergence and robustness of the model. We suggest that the authors provide the loss function curve of the training process to demonstrate the convergence and robustness of the model. (7)There are grammatical errors in the paper. For example, in the fifth page of the paper, the sentence “In contrast, our proposed multi-contexts discriminator provides three kind of supervisions…” uses “three kind of supervisions” instead of “three kinds of supervision.” We suggest that the authors thoroughly check the grammar of their paper. (8)The reference citation order in the paper is mixed up. We suggest that the authors re-sort the “Reference” section according to the order of appearance in the paper.
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

4
Please justify your recommendation. What were the major factors that led you to your overall score for this paper?

The research objective and significance of this paper are relatively clear, but there are significant problems with the description of the model structure, experimental details, and result analysis.
Reviewer confidence

Very 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 proposes the MCGAN, a multi-contexts discriminator based generative adversarial network for predicting diverse FC from SC.
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 question is important to address. The novelty of paper is relative good.
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 comparison results can be enhanced.
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

OK for reproducibility
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/2023/en/REVIEWER-GUIDELINES.html

This paper proposes the MCGAN, a multi-contexts discriminator based generative adversarial network for predicting diverse FC from SC. My suggestion to this paper is listed as follows, 1 in the comparison, the author compared with several methods. But the lastest work of these method are in 2020. More recent works should be mentioned in the introduction and compared here. 2 the movitation of this paper should be enhanced.
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?

See 9
Reviewer confidence

Very 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

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 proposed a novel approach to map brain functional network from brain structural network. The methodology and results are presented clearly. Although one reviewer pointed out some issues, those issues are minor and the main attitude is enthusiastic. Considering the other two reviewers’ comments, I would recommend accept.

Author Feedback

N/A

back to top

Predicting Diverse Functional Connectivity from Structural Connectivity Based on Multi-Contexts Discriminator GAN