Abstract

Resting-state fMRI (rs-fMRI) offers a powerful tool for analyzing functional organization in brain for neurodegenerative and neurodevelopmental disorders. Although graph-based and spatio-temporal models have shown promise, most existing approaches decouple spatial structure from temporal dynamics or rely on predefined temporal windows, limiting the capacity of spatial information to directly modulate temporal processing. To address these limitations, we propose a spectral–temporal state space framework that integrates graph spectral representations with state-space modeling through a spectral-temporal kernel, enabling window-free and frequency-based temporal modeling. Experiments on three rs-fMRI cohorts with 1,926 subjects demonstrate that our method consistently outperforms competitive baselines across diverse conditions, including Parkinson’s disease, autism spectrum disorder, and attention deficit hyperactivity disorder. In addition, we provide interpretable insights into spectral–temporal patterns of brain dysfunction, advancing the characterization and classification of neurodegenerative and neurodevelopmental disorders.

Links to Paper and Supplementary Materials

Main Paper (Open Access Version): https://papers.miccai.org/miccai-2026/paper/0830_paper.pdf

SharedIt Link: Not yet available

SpringerLink (DOI): Not yet available

Supplementary Material: Not Submitted

Link to the Code Repository

N/A

Link to the Dataset(s)

N/A

BibTex

@InProceedings{SimJae_SpectralTemporal_MICCAI2026,
        author = { Sim, Jaeyoon AND Lee, Hoseok AND Park, Jihwan AND Baek, Seunghun AND Zhang, Yu AND Kim, Won Hwa},
        title = { { Spectral-Temporal State Space Modeling on Functional Brain Networks } },
        booktitle = {Medical Image Computing and Computer Assisted Intervention -- MICCAI 2026},
        year = {2026},
        publisher = {Springer Nature Switzerland},
        volume = {LNCS 16894},
        month = {September},
        page = {pending}
}


Reviews

Review #1

  • Please describe the contribution of the paper

    The paper introduces a spectral-temporal framework that integrates spectral decomposition of the connectivity matrix with Mamba to classification of functional MRI scans to healthy and disease classes.

  • Please list the major strengths of the paper: you should highlight 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.Though spectral-Mamba combination has been utilized in other domains, this may be the first time that applies to fMRI analysis. 2.Experimental results show superior performance in disease detection task

  • Please list the major weaknesses of the paper. Please provide details: for instance, if you state that a formulation, way of using data, demonstration of clinical feasibility, or application is not novel, then you must provide specific references to prior work.

    1.To account for site-specific effects, authors resort to chop time-series to match to smallest one. This must have amplified the performance of their methods and lead to unfair comparison with the existing methods. 2.The performance improvements are marginal and whether they are statistically significantly better with the existing methods is unclear. 3.Dynamic functional connectivity assume the connectivity matrix itself is temporally changing but the current method assumes the connectivity matrix is constant across the experiment and temporal changes are attributed to others.

  • 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.

  • Based on your review and your understanding of the MICCAI Scientific Code of Ethics, do you believe this submission may involve a potential ethics concern or violation?

    N/A

  • Optional: If you have any additional comments to share with the authors, please provide them here. Please also refer to our Reviewer’s guide on what makes a good review and pay specific attention to the different assessment criteria for the different paper categories: https://conferences.miccai.org/2026/en/REVIEWER-GUIDELINES.html

    N/A

  • 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.

    (4) Weak Accept — marginally above the acceptance threshold, but would not mind if rejected, dependent on rebuttal

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

    See the strengths and weaknesses above.

  • Reviewer confidence

    Very confident (4)

  • [Post rebuttal] After reading the authors’ rebuttal, please state your final opinion of the paper.

    N/A

  • [Post rebuttal] Please justify your final decision from above.

    N/A



Review #2

  • Please describe the contribution of the paper

    This paper proposes a spectral-temporal framework for functional brain networks. The core idea is to tightly couple graph spectral representations with state-space temporal modeling. This method projects fMRI signals onto a graph spectral basis derived from the Laplacian and introduces a spectral–temporal kernel that jointly controls amplitude modulation and temporal decay. The kernel is reformulated as an SSM compatible with the Mamba architecture. The window-free temporal modeling can further avoid sliding-window heuristics. Experiments on three benchmarks (PPMI, ABIDE, ADHD-200; total 1,926 subjects) show the advantages of the proposed method.

  • Please list the major strengths of the paper: you should highlight 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 idea of combining graph spectral theory and SSMs is interesting. 2.The methodology is clearly presented and has solid mathematical groundings. 3.The method exhibits consistent performance improvements across three datasets.

  • Please list the major weaknesses of the paper. Please provide details: for instance, if you state that a formulation, way of using data, demonstration of clinical feasibility, or application is not novel, then you must provide specific references to prior work.

    1.Sections 4 and 5 should be put into the same section. 2.The methodology part could be more clearly explained. For example, the spectral-temporal kernel intuition is mathematically clear but not sufficiently explained conceptually. 3.The ablation results could contain more variants, e.g., different spectral filters and different spectral components.

  • Please rate the clarity and organization of this paper

    Satisfactory

  • 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.

  • Based on your review and your understanding of the MICCAI Scientific Code of Ethics, do you believe this submission may involve a potential ethics concern or violation?

    N/A

  • Optional: If you have any additional comments to share with the authors, please provide them here. Please also refer to our Reviewer’s guide on what makes a good review and pay specific attention to the different assessment criteria for the different paper categories: https://conferences.miccai.org/2026/en/REVIEWER-GUIDELINES.html

    N/A

  • 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.

    (4) Weak Accept — marginally above the acceptance threshold, but would not mind if rejected, dependent on rebuttal

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

    See strengths

  • Reviewer confidence

    Very confident (4)

  • [Post rebuttal] After reading the authors’ rebuttal, please state your final opinion of the paper.

    N/A

  • [Post rebuttal] Please justify your final decision from above.

    N/A



Review #3

  • Please describe the contribution of the paper

    The manuscript proposes SpecT-Mamba, a spectral-temporal state space framework for classifying neurological and neurodevelopmental disorders from resting-state fMRI (rs-fMRI) data.

  • Please list the major strengths of the paper: you should highlight 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 project BOLD time series onto the graph spectral domain via truncated graph eigenvectors, then apply a learnable spectral-temporal kernel within a state-space model (SSM) framework in which the temporal decay rate of each spectral channel is conditioned on the corresponding graph eigenvalue. This kernel is reformulated as a Mamba block, enabling efficient causal sequential modeling. Experiments are conducted on three public rs-fMRI benchmarks—PPMI (181 subjects), ABIDE (989 subjects), and ADHD-200 (776 subjects)—against 15 static and spatio-temporal baselines, with additional interpretability analysis via Grad-CAM.

  • Please list the major weaknesses of the paper. Please provide details: for instance, if you state that a formulation, way of using data, demonstration of clinical feasibility, or application is not novel, then you must provide specific references to prior work.

    Eq. (5) indicates that the BOLD signals are predictable. This violates the unpredictable and fluctuating nature of rs-fMRI signals. The authors should provide justification for this assumption.

  • Please rate the clarity and organization of this paper

    Satisfactory

  • 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.

  • Based on your review and your understanding of the MICCAI Scientific Code of Ethics, do you believe this submission may involve a potential ethics concern or violation?

    N/A

  • Optional: If you have any additional comments to share with the authors, please provide them here. Please also refer to our Reviewer’s guide on what makes a good review and pay specific attention to the different assessment criteria for the different paper categories: https://conferences.miccai.org/2026/en/REVIEWER-GUIDELINES.html

    N/A

  • 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.

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

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

    The work is supported by extensive experimental validation and rigorous mathematical foundations.

  • Reviewer confidence

    Confident but not absolutely certain (3)

  • [Post rebuttal] After reading the authors’ rebuttal, please state your final opinion of the paper.

    N/A

  • [Post rebuttal] Please justify your final decision from above.

    N/A



Author Feedback

We thank all reviewers for their valuable feedback and for recognizing the contributions of our work. Here, we clarify the raised questions and concerns. [Q] Use of chopped time-series (R1) [A] The chopping procedure was not applied exclusively to our method. Instead, all baseline methods were evaluated using the same chopped datasets under identical settings, ensuring a fair comparison. Moreover, the resulting sequence lengths remain sufficiently long for meaningful temporal modeling (112 steps for PPMI, 120 steps for ABIDE, and 72 steps for ADHD-200), indicating that the chopping process does not substantially compromise temporal information. [Q] Marginal performance improvement (R1) [A] Although the performance improvements are relatively modest, our method is evaluated against strong SOTA baselines, where achieving large gains is inherently challenging. Beyond predictive performance, the key contribution of this work lies in being the first to bridge spectral kernels and state space models within a unified framework. In addition, incorporating this formulation into the Mamba architecture provides favorable computational efficiency alongside competitive performance. [Q] Constant connectivity matrix (R1) [A] Unlike conventional dynamic functional connectivity approaches, our framework is intentionally designed to use a stable connectivity structure together with evolving temporal signals, rather than estimating time-varying connectivity matrices via sliding windows. This design choice allows dynamic temporal information to be captured directly from the underlying signals while avoiding the instability and heuristic hyperparameter sensitivity associated with window-based methods. Consequently, our approach enables effective window-free spatio-temporal modeling while preserving meaningful temporal dynamics. [Q] Improve clarity (R2) [A] We appreciate the suggestion and will incorporate the feedback to improve the clarity and organization of the paper in the camera-ready version. [Q] Additional ablation studies (R2) [A] We appreciate the suggestion. However, we would like to clarify that the proposed spectral-temporal kernel itself constitutes the core component of our method, rather than a replaceable filtering module. Therefore, there are no alternative spectral filters within our framework to substitute directly. Regarding the spectral components, we indeed experimented with different values of D through hyperparameter tuning and selected the best-performing configurations for the reported results. Due to space limitations, we were unable to include the full set of ablation results, but we will clarify this more explicitly in the camera-ready version. [Q] Predictability of BOLD signal (R3) [A] We believe the reviewer may have interpreted Eq. (5) as imposing deterministic dynamics due to its exponential form. However, our formulation does not assume that rs-fMRI BOLD signals are fully predictable. Instead, it models statistical temporal dependencies. Despite their noisy and fluctuating nature, rs-fMRI signals exhibit structured temporal correlations, which our spectral-temporal kernel is designed to capture without enforcing strict predictability. Similar assumptions are widely adopted in prior rs-fMRI studies, including sliding-window and spatio-temporal models. Therefore, Eq. (5) should be understood as a filtering mechanism within a state-space framework, rather than a deterministic generative model.




Meta-Review

Meta-review #1

  • Your recommendation

    Provisional Accept

  • Please justify your decision. 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.

    All reviewers have rated the submission favourably and recommended acceptance.



back to top