List of Papers By topics Author List
| Paper Info | Reviews | Meta-review | Author Feedback | Post-Rebuttal Meta-reviews |
Authors
Jiale Cheng, Xin Zhang, Fenqiang Zhao, Zhengwang Wu, Xinrui Yuan, Li Wang, Weili Lin, Gang Li
Abstract
Exploring the relationship between the cognitive ability and infant cortical structural and functional development is critically important to advance our understanding of early brain development, which, however, is very challenging due to the complex and dynamic brain development in early postnatal stages. Conventional approaches typically use either the structural MRI or resting-state functional MRI and rely on the region-level features or inter-region connectivity features after cortical parcellation for predicting cognitive scores. However, these methods have two major issues: 1) spatial information loss, which discards the critical fine-grained spatial patterns containing rich information related to cognitive development; 2) modality information loss, which ignores the complementary information and the interaction between the structural and functional images. To address these issues, we unprecedentedly invent a novel framework, namely cortical surface-based multimodal learning framework (CSML), to leverage fine-grained multimodal features for cognition development prediction. First, we introduce the fine-grained surface-based data representation to capture spatially detailed structural and functional information. Then, a dual-branch network is proposed to extract the discriminative features for each modality respectively and further captures the modality-shared and complementary information with a disentanglement strategy. Finally, an age-guided cognition prediction module is developed based on the prior that the cognition develops along with age. We validate our method on an infant multimodal MRI dataset with 318 scans. Compared to state-of-the-art methods, our method consistently achieves superior performances, and for the first time suggests crucial regions and features for cognition development hidden in the fine-grained spatial details of cortical structure and function.
Link to paper
DOI: https://doi.org/10.1007/978-3-031-43895-0_58
SharedIt: https://rdcu.be/dnwzq
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 disentangled multi-modal learning framework for predicting infant cognitive development. The proposed approach incorporates multi-view features from various imaging modalities and employs a disentanglement strategy to extract age-independent cognitive features. The model is evaluated on the UNC/UMN Baby Connectome Project dataset and achieves the good performance.
- 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 biggest innovation of this paper is the first-time utilization of a large-scale dataset of healthy infants to predict cognitive abilities, achieving a significant improvement in prediction accuracy compared to the existing algorithms. The model design innovation lies in the utilization of a multi-task prediction approach that considers both age and cognitive abilities, thus considering the robust correlation between these two factors in infants. The paper is well-organized and has a relatively good readability.
- 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 paper does not provide evaluation results of the model’s age prediction ability. 2)The author did not conduct an ablation study to compare the differences between the multi-task prediction (age and cognitive ability) and single-task prediction (cognitive ability) based models.
- 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
I believe that the obtained results can, in principle, be reproduced.
- 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）It is recommended to provide an evaluation result of age prediction for this model to determine whether it has accurately learned age-related information. 2）It is suggested to add ablation experiments to directly predict cognitive scores using multimodal features and compare the results with those derived from multi-task based behavior prediction model, verifying the necessity and importance of simultaneously predicting age and behavior in the paper. 3）The superscript in the second term of Equation 11 should be changed from 1 to 2.
- 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?
1) The manuscript is well-organized and well-written. 2) This is the first study to utilize multimodal neuroimaging data for predicting cognitive development. 3) The proposed method exhibits good behavioral prediction ability.
- 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
To address the issues of previous methods using the spatial information loss or the modality information loss, this paper proposes a novel framework, cortical surface-based multimodal learning framework (CSML), to leverage fine-grained multimodal features for cognition development prediction. Experiments on an infant multi-modal MRI dataset with 318 scans demonstrate the superior performance over state-of-the-art methods.
- 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 issues are significant.
- The idea is novel and very clear.
- This paper is well written.
- Good experimental results and analysis.
- 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 methods are relatively few and old. There are only four comparative methods used. Additionally, the latest method was published in 2020.
-
Only one dataset is used, which might restrict the impact of the proposed framework.
-
It is better to point out the future work.
-
- 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 proposed method can be reproduced.
- 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
To address the issues of previous methods using the spatial information loss or the modality information loss, this paper proposes a novel framework, cortical surface-based multimodal learning framework (CSML), to leverage fine-grained multimodal features for cognition development prediction. Experiments on an infant multi-modal MRI dataset with 318 scans demonstrate the superior performance over state-of-the-art methods.
This paper is well written, and its idea is novel and clear, but i still have several questions as follows:
-
The abstract is too long for a conference paper.
-
It is better to add more comparison methods, especially the latest ones.
-
If possible, please add a table to show the training procedure, which can make the method clearer.
-
\lambda_1, \lambda_2 and \lambda_3 might can be reduced to \lambda_1 and \lambda_2, since \lambda_3=1. Additionally, it should have a major loss during the three terms in Eq. (11).
-
There are some typos, like Q_s()->Q_s(\cdot), K_s()->K_s(\cdot), U_s()-> U_s(\cdot), W_s()->W_s(\cdot).
-
- 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?
The issues are significant and this paper clearly show how to solve them.
- 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
The paper proposed a dual-branch surface network to simultaneously extract structural morphologic features and functional connectivity features on cortical surfaces, and further fuse their complementary information in a feature disentanglement module.
- 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.
Considering that cognitive functions develop rapidly during the first years of life, the regressor would be prone to learn the age-related information instead and thus cannot differentiate the individualized development discrepancy between subjects within the same age group. Therefore, the paper used the identify-related feature to predict the cognitive scores under the guidance of the corresponding age feature.
- 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) No ablation experiments were performed to demonstrate the effectiveness of each module. (2) The methods used for comparisons should include more latest prediction models proposed in recent years.
- 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
There is no code released.
- 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) Ablation experiments are recommended to demonstrate the effectiveness of each module, for example, we can take sMRI and fMRI as inputs respectively, then observe predictive outcomes to demonstrate that the use of multimodal images as inputs does improve predictive accuracy. (2) A single-task experiment could be added, which only predicted the cognitive scores but not the age at the same time, and the results were compared with the results of the joint task, to prove the validity of age information to improve the prediction of cognitive scores. (3) The methods used for comparisons should include more latest prediction models proposed in recent years.
- 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?
Compared with KNN baseline method and graph convolution method, the proposed method can get good prediction results of the cognitive scores in this paper, and the lowest RMSE and the highest PCC are obtained, which prove the superiority of the proposed method. However, the results obtained using this method were not optimal compared with those from other cognitive score prediction articles. To the best of my knowledge, a prediction of cognitive scores based on cortical features in one paper yielded an RMSE of 0.023. Therefore, the methodology in this article needs to be further improved to obtain better performance.
- 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.
There is consensus among the reviewers that this is a good paper with more strengths than weaknesses. The presented work addresses a problem of interest for the neuroimaging community in a novel way. The methodology is sound and is evaluated in a relatively large cohort, achieving good performance. Overall, the paper is well-written and few weaknesses are noted. These include the lack of ablation studies as well as the lack of comparisons with single-task models. I would add that the lack of evaluation of model performance in an independent dataset is also a limitation. I would recommend to add a link for the code and share it in order to ensure that the research is reproducible.
Author Feedback
N/A