Paper Info Reviews Author Feedback Meta-Review Back to top
List of Papers Browse by Subject Areas Author List

Abstract

Individualized brain parcellations derived from functional MRI (fMRI) are essential for discerning unique functional patterns of individuals, facilitat-ing personalized diagnoses and treatments. Unfortunately, as fMRI signals are inherently noisy, establishing reliable individualized parcellations typ-ically necessitates long-duration fMRI scan (> 25 min), posing a major chal-lenge and resulting in the exclusion of numerous short-duration fMRI scans from individualized studies. To address this issue, we develop a novel Consecutive-Contrastive Spherical U-net (CC-SUnet) to enable the predic-tion of reliable individualized brain parcellation using short-duration fMRI data, greatly expanding its practical applicability. Specifically, 1) the wide-ly used functional diffusion map (DM), obtained from functional connec-tivity, is carefully selected as the predictive feature, for its advantage in tracing the transitions between regions while reducing noise. To ensure a robust depiction of brain network, we propose a dual-task model to predict DM and cortical parcellation simultaneously, fully utilizing their reciprocal relationship. 2) By constructing a stepwise dataset to capture the gradual changes of DM over increasing scan durations, a consecutive prediction framework is designed to realize the prediction from short-to-long gradual-ly. 3) A stepwise-denoising-prediction module is further proposed. The noise representations are separated and replaced by the latent representa-tions of a group-level diffusion map, realizing informative guidance and de-noising concurrently. 4) Additionally, an N-pair contrastive loss is intro-duced to strengthen the discriminability of the individualized parcella-tions. Extensive experimental results demonstrated the superiority of our proposed CC-SUnet in enhancing the reliability of the individualized par-cellation with short-duration fMRI data, thereby significantly boosting their utility in individualized studies.

Links to Paper and Supplementary Materials

Main Paper (Open Access Version): https://papers.miccai.org/miccai-2024/paper/3653_paper.pdf

SharedIt Link: https://rdcu.be/dV1Ms

SpringerLink (DOI): https://doi.org/10.1007/978-3-031-72069-7_9

Supplementary Material: N/A

Link to the Code Repository

N/A

Link to the Dataset(s)

N/A

BibTex

@InProceedings{Hu_ConsecutiveContrastive_MICCAI2024,
        author = { Hu, Dan and Han, Kangfu and Cheng, Jiale and Li, Gang},
        title = { { Consecutive-Contrastive Spherical U-net: Enhancing Reliability of Individualized Functional Brain Parcellation for Short-duration fMRI Scans } },
        booktitle = {proceedings of Medical Image Computing and Computer Assisted Intervention -- MICCAI 2024},
        year = {2024},
        publisher = {Springer Nature Switzerland},
        volume = {LNCS 15002},
        month = {October},
        page = {88 -- 98}
}

Reviews

Review #1

  • Please describe the contribution of the paper

    The paper proposed a novel neural network architecture and training losses to generate individual functional parcellation with short-duration rsfMRI data.

  • 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 addressed an important but underexplored problem. Reduce the length of rsfMRI acquisition without compromising the derived FC is crucial for a range of clinical and research applications. 
    • The proposed method is novel, especially the design of consecutive prediction and the usage of contrastive loss to achieve good test-retest similarity and inter-subject variability.
    • The evaluation metrics were comprehensive. Validity, test-retest reliability, inter-subject variability, and homogeneity were evaluated.
  • 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 proposed method relies on ground-truth individual parcellation, which could be noisy when traditional method is used to generate the parcellation. It limits the capabilities of the proposed neural network. Ideally, an unsupervised loss can be used to do parcellation.
    • A lack of discussion on dynamic FC. Forcing parcellation from a short-duration rsfMRI to be similar to that from a long-duration one inevitably ignore information in dynamic FC, which has been reported in a number of literature (e.g. https://www.ncbi.nlm.nih.gov/pmc/articles/PMC7935036/).
    • Some motivations in network design were not well explained.
  • Please rate the clarity and organization of this paper

    Very 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?

    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
    • The author may want to discuss how dynamic FC can be ignored if the parcellation of a short-duration rsfMRI is forced to be similar to a long-duration one.
    • The author may want to discuss why they chose pre-calculated parcellation, while parcellation is essentially an unsupervised learning process (e.g. clustering).
    • The author may want to explain why latent code z was separated into two parts, noise and info, and noise was replaced with z_{grp}, rather than concatenating z_{grp} to z. Does the inclusion of z_{grp} hurt inter-subject variability?
    • A number of designs were not included in ablation study, such as the learnable mask layer, the joint prediction (the DM branch), the use of z_{grp}, etc.
  • 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

    Accept — should be accepted, independent of rebuttal (5)

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

    The addressed problem is significant and the proposed method is novel and showed good performance. Though there are some weaknesses, the paper is still valuable and interesting to read.

  • 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

    The paper presents a novel approach to improve the reliability of individualized functional brain parcellation, utilizing gradual changes from short to long duration scans and a spherical U-net in the novel stepwise denoising module. The training has an appropriate goal of making the parcellations from shorter vs longer scans more similar. This strategy demonstrates advantage when applied to extremely short scans (as short as 2 minutes). The innovative contribution also includes the denoising step to replace the unadjustable short-scan noise with the group-level diffusion map in the latent space. Individualized fMRI parcellation has been challenging for shorter fMRI scans, and it is a common challenge in the study of children and older people with cognitive impairment. Therefore, the proposed method is of clinical interest.

  • 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 proper validity definition measures the similarity between parcellations out of short vs long scans, which is a meaningful optimization target.

    The denoising step strikes an interesting trade-off between realistic individualization and signal-to-noise ratio in parcellation.

    The study offers strong numerical support justifying the design of contrastive loss and the consecutive prediction architecture.

  • 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 validation results using the default parcellation method (None) with MSC dataset (untrained) is better than that with HCP data (same as the trained context). The homogeneity metric is more than doubled. Is there a clear interpretation of this counter-intuitive difference? Is this due to the limitation that the MSC test data contains only ten individuals?

    It is unclear when the validity metric is calculated, what the length of short-duration scans is in HCP data.

  • Please rate the clarity and organization of this paper

    Excellent

  • 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?

    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

    Maybe it is beneficial to consider a larger independent test dataset to acquire more reliable evaluation of the generalizability of the trained models.

  • 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

    Accept — should be accepted, independent of rebuttal (5)

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

    The advantage of the proposed method is sufficiently supported by numerical evidence, which justifies a promising contribution to previously challenging clinical applications (individualized parcelation for extremely short fMRI scans).

  • 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

    The authors trained a spherical unet to enhance the accuracy and reliability of individualized functional brain parcellations from short-duration fMRI scans.

  • 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.

    Increasing the reliability and accuracy of functional brain parcellations from short duration fMRI scans is an important area of focus for both research fMRI, such as studying of inter-individual differences and individualized network targeting for rTMS interventions, and clinical fMRI, such as for pre-surgical brain mapping. The authors offer a novel methodology for training a model to enhance the reliability and accuracy as compared to cleaned fMRI signal and other simpler models.

  • 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 authors did not show a comparison between methods on the MSC dataset. The MSC dataset predictions were limited to 2 minutes rather than the 5 minutes. It would be interesting to see the results on 5 minute scans for the MSC dataset and it is unclear why the authors chose the 2 minute scans rather than 5 minute scans.

  • Please rate the clarity and organization of this paper

    Very 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?

    Providing source code would greatly enhance model and training reproducibility.

  • 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

    The figures are too small when printed to evaluate.

  • 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

    Accept — should be accepted, independent of rebuttal (5)

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

    The domain is of importance to research and clinical fMRI use, the solution appears to be novel, and the authors demonstrate the effectiveness and superiority of their methodology.

  • Reviewer confidence

    Confident but not absolutely certain (3)

  • [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.



back to top