List of Papers Browse by Subject Areas Author List
Abstract
Resting-state functional magnetic resonance imaging (rs-fMRI) serves as a potent means to quantify brain functional connectivity (FC), which holds potential in diagnosing diseases. However, conventional FC measures may fall short in encapsulating the intricate functional dynamics of the brain; for instance, FC computed via Pearson correlation merely captures linear statistical dependencies among signals from different brain regions. In this study, we propose an affinity learning framework for modeling FC, leveraging a pre-training model to discern informative function representation among brain regions. Specifically, we employ randomly sampled patches and encode them to generate region embeddings, which are subsequently utilized by the proposed affinity learning module to deduce function representation between any pair of regions via an affinity encoder and a signal reconstruction decoder. Moreover, we integrate supervision from large language model (LLM) to incorporate prior brain function knowledge. We evaluate the efficacy of our framework across two datasets. The results from downstream brain disease diagnosis tasks underscore the effectiveness and generalizability of the acquired function representation. In summary, our approach furnishes a novel perspective on brain function representation in connectomics. Our code is available at https://github.com/mjliu2020/ALBFR.
Links to Paper and Supplementary Materials
Main Paper (Open Access Version): https://papers.miccai.org/miccai-2024/paper/0803_paper.pdf
SharedIt Link: https://rdcu.be/dV1Ml
SpringerLink (DOI): https://doi.org/10.1007/978-3-031-72069-7_2
Supplementary Material: https://papers.miccai.org/miccai-2024/supp/0803_supp.pdf
Link to the Code Repository
N/A
Link to the Dataset(s)
N/A
BibTex
@InProceedings{Liu_Affinity_MICCAI2024,
author = { Liu, Mengjun and Song, Zhiyun and Chen, Dongdong and Wang, Xin and Zhuang, Zixu and Fei, Manman and Zhang, Lichi and Wang, Qian},
title = { { Affinity Learning Based Brain Function Representation for Disease Diagnosis } },
booktitle = {proceedings of Medical Image Computing and Computer Assisted Intervention -- MICCAI 2024},
year = {2024},
publisher = {Springer Nature Switzerland},
volume = {LNCS 15002},
month = {October},
page = {14 -- 23}
}
Reviews
Review #1
- Please describe the contribution of the paper
The paper presents a novel approach for learning brain function representations using affinity learning and LLM-guided supervision. Despite some areas that need improvement, the proposed method has the potential to advance the field of brain connectivity analysis and disease diagnosis.
- 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 approach: The paper introduces a affinity learning framework for learning informative brain function representations from rs-fMRI data. The proposed method goes beyond the limitations of conventional approaches, such as Pearson correlation coefficient, by capturing complex relationships between brain regions.
- Integration of prior knowledge: The authors propose an innovative idea of incorporating prior knowledge from large language models (LLMs) to guide the affinity learning process.
- Promising results: The experimental results on two datasets (ABIDE and ADNI) for ASD and MCI diagnosis demonstrate the effectiveness of the proposed method. The learned function representations show competitive performance compared to state-of-the-art methods, indicating their potential for brain disease diagnosis.
- Generalizability: The transfer learning study suggests that the pre-trained affinity learning model can be applied to unseen datasets without fine-tuning, demonstrating the generality of the proposed approach.
- 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.
Methodological clarity: While the overall idea of affinity learning is promising, the description of the methodology lacks some clarity. The authors should provide more details on how the affinity encoder and signal reconstruction decoder work together to generate the function representations, is it just a auto encoder with similarity measure? Additionally, a clearer explanation of how the LLM-guided supervision is integrated into the learning process would strengthen the methodological contribution. Ablation studies: The paper would benefit from more detailed ablation studies to investigate the contribution of different components in the proposed framework. • Why is a shared projection layer necessary for a pair of input signals, and what would be the implications of using separate projection layers for each? • For example, analyzing the impact of the LLM-guided supervision on the learned function representations and the overall performance would provide valuable insights into its effectiveness. • Why is it necessary to utilize a Language Model (LLM), and can domain-specific knowledge contribute to constructing the similarity matrix of ROIs more effectively? Analysis and interpretation: While the paper presents promising results, a deeper analysis of the learned function representations and their biological significance would enhance the contribution. The authors should provide more insights into how the learned representations capture meaningful brain functional connectivity patterns and discuss their findings in the context of existing literature. Presentation and organization: The paper would benefit from a thorough proofreading to address typographical and grammatical errors. Improving the structure and organization of the paper. Typographical and grammatical errors: o Example: In the “Introduction” section, the authors mention “Brian function representation” instead of “Brain function representation.” Even your key words include typo. Figure and table captions: o Example: In Figure 1, the caption “The architecture of the proposed affinity learning based brain function representation method and its application in the diagnosis of brain diseases” is quite lengthy and lacks clarity.
- Please rate the clarity and organization of this paper
Good
- Please comment on the reproducibility of the paper. Please be aware that providing code and data is a plus, but not a requirement for acceptance.
The authors claimed to release the source code and/or dataset upon acceptance of the submission.
- Do you have any additional comments regarding the paper’s reproducibility?
N/A
- Please provide detailed and constructive comments for the authors. Please also refer to our Reviewer’s guide on what makes a good review. Pay specific attention to the different assessment criteria for the different paper categories (MIC, CAI, Clinical Translation of Methodology, Health Equity): https://conferences.miccai.org/2024/en/REVIEWER-GUIDELINES.html
Please refer to the weakness.
- Rate the paper on a scale of 1-6, 6 being the strongest (6-4: accept; 3-1: reject). Please use the entire range of the distribution. Spreading the score helps create a distribution for decision-making
Weak Accept — could be accepted, dependent on rebuttal (4)
- Please justify your recommendation. What were the major factors that led you to your overall score for this paper?
The approach is novel and the results are promising.
- Reviewer confidence
Very confident (4)
- [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
This paper proposes a new affinity learning method under the guide of LLM models. Experiments are comprehensive.
- 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.
This paper proposes a new affinity learning method under the guide of LLM models. Experiments are comprehensive.
- 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 use of LLM needs expensive cost as well as prior knowledge. For example, you need prior knowledge to input text to ChatGPT.
- Please rate the clarity and organization of this paper
Good
- Please comment on the reproducibility of the paper. Please be aware that providing code and data is a plus, but not a requirement for acceptance.
The submission does not mention open access to source code or data but provides a clear and detailed description of the algorithm to ensure reproducibility.
- Do you have any additional comments regarding the paper’s reproducibility?
No source codes were provided.
- Please provide detailed and constructive comments for the authors. Please also refer to our Reviewer’s guide on what makes a good review. Pay specific attention to the different assessment criteria for the different paper categories (MIC, CAI, Clinical Translation of Methodology, Health Equity): https://conferences.miccai.org/2024/en/REVIEWER-GUIDELINES.html
Positive: This paper proposes a new affinity learning method under the guide of LLM models. Experiments are comprehensive. Negative: The use of LLM needs expensive cost as well as prior knowledge. For example, you need prior knowledge to input text to ChatGPT.
- Rate the paper on a scale of 1-6, 6 being the strongest (6-4: accept; 3-1: reject). Please use the entire range of the distribution. Spreading the score helps create a distribution for decision-making
Weak Accept — could be accepted, dependent on rebuttal (4)
- Please justify your recommendation. What were the major factors that led you to your overall score for this paper?
The novelty is good enough and experiments are comprehensive. However, no source code supports the reproducibility. Hence, it is a boardline paper.
- Reviewer confidence
Very confident (4)
- [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 study proposed an affinity learning framework for modeling FC, leveraging a pre-training model to discern informative function representation among brain regions.
- 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 affinity learning for the function representation is novel in this work.
- 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 introduction of the ChatGPT.
- Please rate the clarity and organization of this paper
Good
- Please comment on the reproducibility of the paper. Please be aware that providing code and data is a plus, but not a requirement for acceptance.
The authors claimed to release the source code and/or dataset upon acceptance of the submission.
- Do you have any additional comments regarding the paper’s reproducibility?
N/A
- Please provide detailed and constructive comments for the authors. Please also refer to our Reviewer’s guide on what makes a good review. Pay specific attention to the different assessment criteria for the different paper categories (MIC, CAI, Clinical Translation of Methodology, Health Equity): https://conferences.miccai.org/2024/en/REVIEWER-GUIDELINES.html
I. The abstract points out the conventional FC measures may fall short in encapsulating the intricate functional dynamics of the brain. After introducing the novel affinity learning framework, it fails to explicitly state the advantages of this approach over conventional methods or clarify the specific problems it addresses. Where in the new framework is dynamic information exemplified? II. In Section 2.1, “Patch Extraction” specifies that three-dimensional patches are randomly sampled from gray matter, with the mean gray matter signal within each patch serving as the functional signal for that area. This contradicts Figure 1, which presents “Mean Time Series of Any Two ROIs”. III. The method in Section 2.3, line 4, which utilizes ChatGPT to assess inter-regional correlations, is deemed unreliable. Employing ChatGPT to introduce prior knowledge is unreliable. IV. In Section 2.2, Equation 2 contains the function ‘tanh’ without providing an explanation. V. Figure 2 is of poor clarity. VI. In Subsection 2.4, The Process in which the input matrix is compressed through the FNN network is not explicated in detail. Ⅶ. In Section 7, Subsection 2.2, regarding the Signal Reconstruction Decoder, the explanation of the reference signal and the target signal for reconstruction is insufficiently clear. Ⅷ. In Figure 1, the process of inputting rij and si into the Decoder to obtain sj, and then inputting rij and sj into the Decoder to retrieve si, is not elaborated upon in Section 2.2 concerning the Signal Reconstruction Encoder, nor is it elucidated within Equation (3).
- Rate the paper on a scale of 1-6, 6 being the strongest (6-4: accept; 3-1: reject). Please use the entire range of the distribution. Spreading the score helps create a distribution for decision-making
Weak Accept — could be accepted, dependent on rebuttal (4)
- Please justify your recommendation. What were the major factors that led you to your overall score for this paper?
The framewok of this paper is interesting.
- Reviewer confidence
Very confident (4)
- [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
Author Feedback
N/A
Meta-Review
Meta-review not available, early accepted paper.