List of Papers By topics Author List
| Paper Info | Reviews | Meta-review | Author Feedback | Post-Rebuttal Meta-reviews |
Authors
Lu Zhang, Saiyang Na, Tianming Liu, Dajiang Zhu, Junzhou Huang
Abstract
Multimodal fusion of different types of neural image data offers an invaluable opportunity to leverage complementary cross-modal information and has greatly advanced our understanding of mild cognitive impairment (MCI), a precursor to Alzheimer’s disease (AD). Current multi-modal fusion methods assume that both brain’s natural geometry and the related feature embeddings are in Euclidean space. However, recent studies have suggested that non-Euclidean hyperbolic space may provide a more accurate interpretation of brain connectomes than Euclidean space. In light of these findings, we propose a novel graph-based hyperbolic deep model with a learnable topology to integrate the individual structural network with functional information in hyperbolic space for the MCI/NC (normal control) classification task. We comprehensively compared the classification performance of the proposed model with state-of-the-art methods and analyzed the feature representation in hyperbolic space and its Euclidean counterparts. The results demonstrate the superiority of the proposed model in both feature representation and classification performance, highlighting the advantages of using hyperbolic space for multimodal fusion in the study of brain diseases. (Code is available here3.)
Link to paper
DOI: https://doi.org/10.1007/978-3-031-43904-9_65
SharedIt: https://rdcu.be/dnwIb
Link to the code repository
https://github.com/nasyxx/MDF-HS
https://github.com/qidianzl/Multimodal-Deep-Fusion-in-Hyperbolic-Space
Link to the dataset(s)
N/A
Reviews
Review #2
- Please describe the contribution of the paper
A novel graph-based hyperbolic deep model with a learnable topology to integrate the individual structural network with functional information in hyperbolic space is proposed in the paper. Motivated by the recent advances that suggest that non-Euclidean hyperbolic space may provide a more accurate interpretation of brain connectomes, the work has been conducted.
- 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.
Well motivated and a good problem formulation. Data has been accumulated methodically and all the details are presented.
- 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 results are not too impressive although Table 2 is convincing. Figure 1 is not explained well.
- 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 it is reproducible. The supllementary material has also been provided.
- 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
Readability of the paper is not particularly great and it took a while to understand the contribution. In related work, it is important to cover the domain details and not just the mathematical approach. It has to be expanded.
- 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 entire thought process is very refreshing. A thorough treatment provided helps.
- 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 #5
- Please describe the contribution of the paper
This paper1 proposes a novel graph-based hyperbolic deep model with a learnable topology to integrate the individual structural network with functional information in hyperbolic space for the MCI/NC (normal control) classification task. The paper shows that the proposed model outperforms state-of-the-art methods in both feature representation and classification performance1.
- 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 is well-written and easy to follow
- the overall idea is interesting and less-explored by the current study
- 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.
authors should provide more figures to illustrate their idea. plus the motivation is not clear enough, authors should explain with more details why using GCN in this specific task. In the experiments, authors should compare with recent SOTA method, like CG-GCN[1]
[1] An Efficient Person Clustering Algorithm for Open Checkout-free Groceries ECCV 2022
- 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
YES
- 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
address the concerns in point 6
- 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?
overall well-written paper with interesting idea, but need to supply more details about the motivation, and comparison.
- 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 #4
- Please describe the contribution of the paper
The paper proposes a new method for Mild Cognitive Impairment Study. Like in the literatures, the graph data with tree-like structure lie in the non-Euclidean space and embedding in Euclidean space will lead to large embedding distortion. Therefore, the authors turn to Hyperbolic space with negative curvature and expect to benefit from both feature learning and classification. Specifically, a Hyperbolic Graph Convolutional network (HGCN) is applied to perform information aggregation/fusion for different modalities. The method is new for this study and the experimental results is convincing.
- 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.
Introducing Hyperbolic Neural Network into Mild Cognitive Impairment is novel and interesting. The paper is also well-organized and easy to follow. The experimental setting is reasonable and the results are convincing.
- 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.
Hyperbolic Graph neural network (HGCN) is relatively old and there are also some space to improve.
- 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
Code related to this workwill release if this work is accepted
- 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
-
I personally think what can further be improved is the method. HGCN is relatively old, where the information fusion is done with the help of tangent space. Since this work uses a tangent space (map between manifold and such vector space), can you explain why such model can benefit from such space, when compared to Euclidean space?
-
Can the authors explain why the feature embedding of hyperbolic space is close to the boundary(Fig. 1 b)? From the visualization, seems the Euclidean space is also not bad, is this correct?
-
- 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?
Like above
- 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 #1
- Please describe the contribution of the paper
This paper presents a novel graph-based hyperbolic deep model and uses that to integrate multimodal (structural and functional) information for MCI/NC classification task. Comprehensive experiments were performed to validate the proposed hyperbolic model with state-of-the-art methods and analyze feature representation in hyperbolic space and its the Euclidean counterparts. The results also showed the superiority.
- 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.
1) the projection of functional and structural connections into non-Euclidean space (hyperbolic space) separately for conventional operations; 2) improvement of the HGCN model for classification in hyperbolic space.
- 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) in Equation 7, why did calculate the functional connectivity matrix by Gaussian kernel rather than Pearson’s correlation? In general, functional connectivity is obtained by the Pearson correlation coefficient, which contains positive and negative values, and some researchers (PMID: 18976716, PMID: 24128734) only included the positive correlation, since the negative correlation is considered spurious connectivity. 2) the statistics of structurally connected fiber bundles are not known whether the authors obtained them by deterministic tracing methods or probabilistic tracing methods. If the deterministic tracing method is used, it is possible that the number of fiber tract statistics from one brain region to others is zero, thus appearing as isolated brain regions. The tracing method needs to be indicated as well as the parameters. 3) In the data setup in Section 4.1, the authors simply used 5-fold cross-validation for one time, which decreased the reliability of the results, authors should consider more times to guarantee the reliability of the results. 4) In the comparison of classification performance in Section 4.2, the authors simply re-implemented Zhang et al. (2021, PMID: 34004495) model on their own data and found its accuracy to be lower than that proposed by the authors, but the authors did not apply the same data to other models, especially for the model of Li et al. (2020, PMID: 32112678), to evaluate the performance the model. 5) The whole paper only includes one figure; the current form is not helpful to understand the idea. The paper presentation needs refinement. This paper details the state-of-the-art methods, theory, proposed idea and experiments in text. But it is heavily lack of enough visual illustrations for theoretical concept, network structure, brain mapping, etc. And there is no description about the computing efficiency.
- 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
It is possible to reproduce the method.
- 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
Please refer to the weaknesses pointed out in the item 6 for paper refinement.
- 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?
Completeness of algorithmetic and experimental details and demonstratioins.
- 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 paper proposed a new method for Mild Cognitive Impairment Study. A novel graph-based hyperbolic deep model with a learnable topology to integrate the individual structural network with functional information in hyperbolic space was studied in the paper. The paper conducted extensive experiments and the results show improvement with the hyperbolic space adoption. All reviewers favored this paper. The AC recommended the authors revise the paper accordingly for their camera ready submission.
Author Feedback
N/A