Pith. sign in

REVIEW 4 major objections 7 minor 70 references

Data-Efficient Psychiatric Disorder Detection via Self-supervised Learning on Frequency-enhanced Brain Networks

T0 review · 4 major / 7 minor · reviewed 2026-08-05 · deepseek-v4-flash

Challenge this review Re-run · record.json Download PDF Read on arXiv ↗

Pith's one-line read FENet claims that a self-supervised model jointly learning time-domain and frequency-domain brain-graph views detects ASD and ADHD from fMRI more accurately than existing supervised and self-supervised baselines, especially when only a frac

desk verdict Solid idea, shaky evaluation: the frequency-enhanced SSL method is a real combination, but site mixing and a possible pretraining leak undermine the SOTA and biomarker claims. read the letter →

arxiv 2509.10524 v1 pith:7PN7UGUL submitted 2025-09-04 eess.IV cs.AIcs.LG

Mujie Liu , Mengchu Zhu , Qichao Dong , Ting Dang , Jiangang Ma , Jing Ren , Feng Xia This is my paper · ORCID
classification eess.IVcs.AIcs.LG
keywords psychiatricdisorderdetectionfMRIself-supervisedlearningbrainnetworksfrequency-domainanalysisgraphneuralcanonicalcorrelationdataefficiency
verification ladder T0 review T1 audit T2 compute T3 formal

The pith

A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.

The reading

The paper is trying to establish that fMRI-based psychiatric disorder detection can be made data-efficient by explicitly teaching a model to use both temporal and spectral views of the same brain scan. It proposes FENet, a self-supervised framework that builds two brain graphs from one fMRI time series—one in the time domain and one in the graph-frequency domain—and aligns their learned representations with a consistency objective. The authors claim FENet outperforms all compared graph-based supervised and self-supervised methods on the ABIDE and ADHD-200 datasets while using only 20% of labels for fine-tuning, and that high-frequency components carry a particularly strong disorder signal. If true, this would make fMRI-based screening more practical in clinical settings where labeled psychiatric data are scarce.

What carries the argument

The central mechanism is multi-view brain graph augmentation: the same ROI-level BOLD time series is represented once as a time-domain connectivity graph and once as a graph-frequency graph via the Graph Fourier Transform, sharing the same adjacency matrix. Two domain-specific encoders process these views—a time-domain GCN and a frequency-domain FGNN that uses a graph filter over Laplacian eigenvalues plus Fourier Graph Operator layers to keep computation log-linear. The learning objective is a CCA-style domain consistency loss: it minimizes the distance between time and frequency representations of the same subject while decorrelating each representation's dimensions, balancing alignment wi

What would settle it

Run FENet with site-hold-out evaluation: train on a subset of acquisition sites and test on held-out sites from ABIDE and ADHD-200. If accuracy drops to near chance or the high-frequency components lose their predictive value, site effects rather than disorder-related frequency patterns carry the reported result. Alternatively, train a linear probe to predict the acquisition site from the learned FENet representation; high site-prediction accuracy would indicate the representation encodes site confounds.

Watch

Extended reading notes

Core claim

FENet integrates time-domain functional connectivity with frequency-domain spectral features by transforming the same BOLD signals through the Graph Fourier Transform, then encoding each view with a dedicated graph neural network. A domain-consistency-guided canonical correlation analysis objective pulls the two representations together while decorrelating each view's feature dimensions, producing a fused representation used for disorder classification. On ABIDE, FENet reports accuracy, AUC, and F1-score gains of 2.7%, 1.4%, and 4.9% over the strongest baseline; on ADHD-200 the gains are 3.4%, 0.7%, and 2.3%. Under 20% labeled data, it reports accuracy improvements of 6.5 and 11.2 percentage

Load-bearing premise

The load-bearing premise is that random stratified five-fold splits of multi-site fMRI data make scanner and acquisition-site differences ignorable, so the measured classification performance reflects the disorder rather than the site.

Editorial extensions

If this is right

  • If FENet's claims hold, psychiatric-disorder classifiers can be trained with a fraction of the labels currently needed, reducing the cost of fMRI-based screening and diagnosis.
  • Frequency-domain views can be derived from the same scan without new data collection, so the approach can be added to existing brain-graph SSL pipelines.
  • The reported high-frequency dominance suggests that future diagnostic models should retain high-frequency graph components rather than treating them as noise.
  • The log-linear complexity of the frequency encoder means the method can scale to higher-resolution brain atlases with more ROIs.
  • The larger gains at 20% labeled data compared with CCA-SSG directly support the paper's data-efficiency claim rather than only its absolute accuracy.

Reading between the lines

Editorial extensions of the paper, not claims the author makes directly.

  • A site-hold-out evaluation, training on some acquisition sites and testing on others, would clarify whether the reported gains reflect disorder-related signal or scanner/site differences; the authors themselves flag harmonization as future work.
  • The same time-frequency consistency objective could transfer to other spectral neuroimaging modalities such as EEG or MEG, but the paper only tests fMRI with the AAL atlas.
  • The ablation showing that mid-frequency components add little suggests a concrete design rule worth testing: drop the mid-band in the frequency encoder on new datasets and check whether accuracy holds.
  • Pretraining on large unlabelled fMRI repositories before fine-tuning could push the minimal-label boundary further; FENet currently pretrains on the same datasets it is evaluated on.
Share X Bluesky LinkedIn Reddit HN

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, and a circularity audit.

Referee Report

4 major / 7 minor

Summary. The paper proposes FENet, a self-supervised framework for fMRI-based detection of psychiatric disorders (ASD and ADHD). It constructs time- and frequency-domain brain graph views from ROI-level BOLD signals, encodes them with separate GNN encoders (TGNN and an FGNN using graph filters and Fourier graph operators), and aligns the two views via a soft-CCA objective with decorrelation terms (Eq. 11). Experiments on ABIDE and ADHD-200 are reported against supervised and SSL baselines, with claims of state-of-the-art accuracy/AUC/F1, data efficiency under 10-20% label rates, and an ablation suggesting that high-frequency components are the most informative for disorder detection.

Significance. If the empirical claims withstand scrutiny, FENet would be a useful contribution: it appears to be the first brain-network SSL method to explicitly combine time- and frequency-domain views under a non-contrastive CCA-style objective, and the FGO-based encoder provides a computationally attractive way to incorporate spectral information. The paper is clearly written, includes thorough ablations of the learning objective and hyperparameters, and provides code links for all baselines. However, the evaluation protocol currently has site-confounding and pretraining-separation issues that make the headline performance and high-frequency biomarker claims uncertain; these are addressable with additional experiments. The paper does not provide code for FENet itself, which limits reproducibility of the proposed method.

major comments (4)
  1. [§5.4 and §5.1] The random stratified 5-fold split is not site-aware. ABIDE aggregates 17 acquisition sites and ADHD-200 aggregates 8 sites, and scanner/protocol/motion effects are strongly site-correlated. With random splits the same sites appear in both training and test folds, so the model can score well by recognizing site-specific artifacts rather than psychiatric disorder. No site harmonization, site-stratified folds, or site-hold-out is reported. This threatens the SOTA claim in §6.1 and the frequency-band interpretation in §6.3.1. Please add a site-hold-out experiment (e.g., leave-site-out) with SSL pretraining confined to training sites, and report per-site or site-group metrics. At minimum, report site composition of folds and demonstrate that the model does not exploit site identity.
  2. [§5.4] The pretraining description states: 'the model is trained in a fully unsupervised manner for 200 iterations on the entire dataset.' If 'entire dataset' includes the test fold, the SSL pretraining stage has transductive access to the test distribution (though not the labels), which can inflate fine-tuning performance—especially in the low-label experiments of §6.2. It must be clarified whether pretraining is restricted to training folds and whether all SSL baselines follow exactly the same train/test separation during pretraining. If the current protocol is transductive, the data-efficiency claim in Figure 7 should be re-run with pretraining on training folds only.
  3. [§6.1, Table 2] No statistical significance tests are reported. Several head-to-head differences are within one standard deviation of the repeated runs: e.g., ABIDE ACC 62.5±3.9 vs GATE 59.8±3.4; ADHD AUC 69.7±4.4 vs A-GCL 69.0±3.2; ADHD F1 67.0±5.2 vs A-GCL 64.7±3.9. Reporting only mean±std is insufficient to support 'FENet outperforms all baseline methods.' Please add paired tests across the same fold/seed structure (e.g., paired bootstrap, McNemar, or corrected resampled t-test) or report confidence intervals.
  4. [§6.3.1 and §5.4] The conclusion that high-frequency components are critical is inferred from the ablation FENet_FH having the best accuracy, but the frequency thresholds λ_L and λ_H were 'optimized' on the same datasets and the final dual-filter architecture was selected after observing these results. This makes the high-frequency biomarker interpretation vulnerable to selection bias. Please pre-specify thresholds from the literature (e.g., [19,44]) or evaluate on an independent hold-out/nested CV, and report the dual-filter model against single-band variants under fixed thresholds.
minor comments (7)
  1. [Algorithm 1, line 3] The initialization line reads 'Initialize: The model parameters θF and θF'; the second should be θT.
  2. [§6.1] The phrase 'recall is slightly lower than A-GCL' understates the ABIDE gap: 63.3±4.8 vs 71.6±12.9. Please quantify the gap and acknowledge the high variance of the A-GCL recall estimate.
  3. [Figure 7] The data-efficiency curves show no error bars or variance information. Given the repeated-seed protocol described in §5.4, please include error bars or state explicitly that the plotted points are means over seeds.
  4. [§4.3.2, Eq. (10)] The definition of S_{0:p} and the role of the complex-valued bias b_p is under-specified. Clarify how complex values are handled before being fed into the real-valued objective in Eq. (11).
  5. [§5.4] '200 iterations' is ambiguous; specify whether this means 200 epochs or 200 optimizer steps.
  6. [Table 1 and §5.4] Table 1 lists K_L and K_H as frequency threshold parameters, but the text uses λ_L and λ_H. Align the notation for consistency.
  7. [Reproducibility] No code release for FENet is mentioned, while links are provided for all baseline methods. Making the proposed method's implementation available would strengthen the reproducibility of the benchmark-style claims.

Circularity Check

1 steps flagged · score 4.0 of 10

Secondary high-frequency importance claim is post-hoc to threshold selection; core benchmark claims remain empirical and non-circular.

  1. fitted input called prediction [Section 5.4 (Implementation Details) and Section 6.3.1 (Ablation Study)]
    "To effectively set the frequency thresholds ... we optimized the λL as the lowest 20% of eigenvalues, while the λH as the top 20% of eigenvalues. To evaluate the performance of both encoders, we tune the hyperparameters in Equation (11) and compare the model classification accuracy ... FENet_FH achieves the best performance on both ABIDE and ADHD datasets, suggesting that high-frequency components often capture rapid neural oscillations ... Motivated by these insights, our approach integrates both low-pass and high-pass filters to extract complementary low- and high-frequency components while"

    The frequency band thresholds (λL, λH) that define 'high-frequency' and the decision to discard mid-frequency components are selected by comparing classification accuracy on the same ABIDE/ADHD data. The later claim that high-frequency components play a critical role is therefore not an independent prediction; it restates which filter configuration won the in-sample model selection. The final dual-filter architecture is justified by the very ablation used to choose it, so the secondary biomarker conclusion reduces to the selection criterion rather than to an out-of-sample test. The headline benchmark comparison (FENet vs. baselines) is empirical and not circular, so the overall circularity is partial and confined to the frequency-importance interpretation.

full rationale

The paper's central performance claim — that FENet outperforms strong SSL baselines on ABIDE and ADHD-200 — is an empirical benchmark result and is not circular: it is a direct comparison of measured accuracy/AUC/F1 under the stated protocol. I found no load-bearing self-citation chain: the author-group references (e.g., [23], [26], [36]) appear only in related-work context and do not justify the method's core design. The only notable circularity is in the secondary 'high-frequency components are critical' claim. The paper optimizes the frequency thresholds and hyperparameters using classification accuracy on the same datasets (Section 5.4), then uses an ablation (FENet_FH best) to conclude that high-frequency information is biologically important (Section 6.3.1). This is a fitted-input-called-prediction pattern: the winning frequency band is selected by the evaluation metric, and the same evaluation is then offered as evidence for the band's importance. However, the central SOTA benchmark does not depend on this interpretation, and the site-mixing / pretraining-on-entire-dataset concerns raised by the skeptic are validity risks, not circularity. I therefore score 4 rather than higher.

Assumptions & free parameters 6 free parameters · 4 assumptions · 0 invented entities

The central claim rests on standard spectral graph theory and CCA, plus domain assumptions about biological plausibility of graph frequencies, alignment of time/frequency views, and unbiasedness of random splits. No new physical entities are introduced; the free parameters are hyperparameters tuned on the same datasets used for evaluation.

free parameters (6)
  • gamma (g) = 1e-5
    Trade-off coefficient for time-domain decorrelation term in Eq. (11), tuned via sensitivity analysis (Figure 9).
  • beta (b) = 1e-4
    Trade-off coefficient for frequency-domain decorrelation term in Eq. (11), tuned via sensitivity analysis (Figure 9).
  • Frequency thresholds lambda_L and lambda_H = lowest 20% and top 20% of Laplacian eigenvalues
    'We optimized the lambda_L as the lowest 20% of eigenvalues, while the lambda_H as the top 20%' (Section 5.4); determines the low/high/mid band split.
  • Number of FGO layers = 3
    Chosen 'based on empirical evaluations' (Section 5.4); affects capacity and the claimed complexity.
  • Edge construction threshold = top 20% positive Pearson correlations
    Defines the edge set E of the brain graphs (Section 4.2.1); a sparsity choice inherited from prior work but still a modeling parameter.
  • Selected frequency component dimension K = not specified
    The filtered feature matrix is said to be N x K with K << N, but K is never given; a free design choice in the FGNN that affects performance.
assumptions (4)
  • domain assumption The graph Laplacian eigendecomposition L = U Lambda U^T yields a biologically meaningful frequency decomposition of BOLD signals.
    Invoked in Section 3.1 and 4.3.2 to justify GFT of brain signals, relying on Huang et al. [15]; not independently validated in this paper.
  • domain assumption Time-domain and frequency-domain latent representations of the same BOLD signal should align in shared latent space (domain consistency).
    Borrowed from Zhang et al. [64] and used to design the CCA objective in Section 4.4; the brain-graph instantiation is new but the principle is assumed.
  • domain assumption Random stratified splits without site harmonization provide unbiased estimates of model performance on ABIDE and ADHD-200.
    The experimental protocol (Section 5.4) assumes site effects are negligible, yet these multi-site datasets are known to have strong site confounds.
  • domain assumption Pearson correlation of BOLD time courses adequately captures functional connectivity, and retaining only the top 20% positive correlations yields a valid brain graph.
    Section 4.2.1 follows prior clinical practice but makes a strong modeling simplification (e.g., discards negative correlations without justification).

how reviews work

0 comments
Cite this review

Pith. "Pith review of Data-Efficient Psychiatric Disorder Detection via Self-supervised Learning on Frequency-enhanced Brain Networks." pith.science (2026). https://pith.science/paper/7PN7UGUL

@misc{pith2026250910524,
  author       = {Pith},
  title        = {Pith review of: Data-Efficient Psychiatric Disorder Detection via Self-supervised Learning on Frequency-enhanced Brain Networks},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/7PN7UGUL}},
  note         = {Machine review of arXiv:2509.10524}
}
read the original abstract

Psychiatric disorders involve complex neural activity changes, with functional magnetic resonance imaging (fMRI) data serving as key diagnostic evidence. However, data scarcity and the diverse nature of fMRI information pose significant challenges. While graph-based self-supervised learning (SSL) methods have shown promise in brain network analysis, they primarily focus on time-domain representations, often overlooking the rich information embedded in the frequency domain. To overcome these limitations, we propose Frequency-Enhanced Network (FENet), a novel SSL framework specially designed for fMRI data that integrates time-domain and frequency-domain information to improve psychiatric disorder detection in small-sample datasets. FENet constructs multi-view brain networks based on the inherent properties of fMRI data, explicitly incorporating frequency information into the learning process of representation. Additionally, it employs domain-specific encoders to capture temporal-spectral characteristics, including an efficient frequency-domain encoder that highlights disease-relevant frequency features. Finally, FENet introduces a domain consistency-guided learning objective, which balances the utilization of diverse information and generates frequency-enhanced brain graph representations. Experiments on two real-world medical datasets demonstrate that FENet outperforms state-of-the-art methods while maintaining strong performance in minimal data conditions. Furthermore, we analyze the correlation between various frequency-domain features and psychiatric disorders, emphasizing the critical role of high-frequency information in disorder detection.

Figures

Figures reproduced from arXiv: 2509.10524 by the authors.

Figure 1
Figure 1. Illustration of fMRI information diversity in psychiatric disorder detection. Brain network analysis of health control [PITH_FULL_IMAGE:figures/full_fig_p003_1.png] view at source ↗
Figure 2
Figure 2. Overview of FENet model. The FENet model consists of three modules: multi-view graph construction, time-frequency [PITH_FULL_IMAGE:figures/full_fig_p008_2.png] view at source ↗
Figure 3
Figure 3. Pipeline of brain graph construction from raw fMRI data. [PITH_FULL_IMAGE:figures/full_fig_p009_3.png] view at source ↗
Figures from the paper (6 more)
Figure 4
Figure 4. Figure 4: The framework of the TGNN module consists of several GCN layers. transform the brain signals XT from the time domain to the spectral domain XF using the GFT operation F𝐺 (·): XF = F𝐺 (XT) = U ⊤XT. This formulation allows us to explicitly introduce frequency information…
Figure 5
Figure 5. Figure 5: The framework of the FGNN module consists of a graph filter and several FGO network layers. [PITH_FULL_IMAGE:figures/full_fig_p011_5.png]
Figure 6
Figure 6. Figure 6: Illustration of the learning objective. For each BOLD signal [PITH_FULL_IMAGE:figures/full_fig_p013_6.png]
Figure 7
Figure 7. Figure 7: Accuracy (%) of FENet and CCA-SSG at different labeled data rates for fine-tuning. [PITH_FULL_IMAGE:figures/full_fig_p018_7.png]
Figure 8
Figure 8. Figure 8: Ablation study performance (%) for different learning objective settings. The vertical lines (error bars) on the graph [PITH_FULL_IMAGE:figures/full_fig_p019_8.png]
Figure 9
Figure 9. Figure 9: Accuracy (%) of FENet with different values of 𝛾 and 𝛽 in Equation (11). of preserving domain-specific semantic structures and avoiding over-alignment, which can otherwise obscure meaningful features unique to each view. Nevertheless, despite these improvements, cosine…

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

70 extracted references · 68 canonical work pages

  1. [1]

    Galen Andrew, Raman Arora, Jeff Bilmes, and Karen Livescu. 2013. Deep canonical correlation analysis. In International Conference on Machine Learning. PMLR, 1247–1255

    work page 2013

  2. [2]

    Owen J Arthurs and Simon Boniface. 2002. How well do we understand the neural origins of the fMRI BOLD signal? TRENDS in Neurosciences 25, 1 (2002), 27–31

    work page 2002

  3. [3]

    Mohammadreza Bayat, Melissa Hernandez, Madeline Curzon, Dea Garic, Paulo Graziano, and Anthony Steven Dick. 2025. Reduced recruitment of inhibitory control regions in very young children with ADHD during a modified Kiddie Continuous Performance Task: a fMRI study. Cortex (2025)

    work page 2025

  4. [4]

    Xia-An Bi, Ke Chen, Siyu Jiang, Sheng Luo, Wenyan Zhou, Zhaoxu Xing, Luyun Xu, Zhengliang Liu, and Tianming Liu. 2023. Community graph convolution neural network for alzheimer's disease classification and pathogenetic factors identification. IEEE Transactions on Neural Networks and Learning Systems (2023)

    work page 2023

  5. [5]

    Defu Cao, Yujing Wang, Juanyong Duan, Ce Zhang, Xia Zhu, Congrui Huang, Yunhai Tong, Bixiong Xu, Jing Bai, Jie Tong, et al. 2020. Spectral temporal graph neural network for multivariate time-series forecasting. Advances in Neural Information Processing Systems 33 (2020), 17766–17778

    work page 2020

  6. [6]

    Xiaobin Chang, Tao Xiang, and Timothy M Hospedales. 2018. Scalable and effective deep CCA via soft decorrelation. In Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition . 1488–1497

    work page 2018

  7. [7]

    Israel Cohen, Yiteng Huang, Jingdong Chen, Jacob Benesty, Jacob Benesty, Jingdong Chen, Yiteng Huang, and Israel Cohen. 2009. Pearson correlation coefficient. Noise Reduction in Speech Processing (2009), 1–4

    work page 2009

  8. [8]

    Hejie Cui, Wei Dai, Yanqiao Zhu, Xuan Kan, Antonio Aodong Chen Gu, Joshua Lukemire, Liang Zhan, Lifang He, Ying Guo, and Carl Yang. 2022. Braingb: A benchmark for brain network analysis with graph neural networks. IEEE Transactions on Medical Imaging 42, 2 (2022), 493–506

    work page 2022

Show all 70 references
  1. [9]

    Junling Du, Shangyu Wang, Rentong Chen, and Shaoping Wang. 2025. Improving fMRI-Based Autism Severity Identification via Brain Network Distance and Adaptive Label Distribution Learning. IEEE Transactions on Neural Systems and Rehabilitation Engineering 33 (2025), 162–174

    2025

  2. [10]

    Yuhui Du, Songke Fang, Xingyu He, and Vince D Calhoun. 2024. A survey of brain functional network extraction methods using fMRI data. Trends in Neurosciences (2024)

    2024

  3. [11]

    Nicha C Dvornek, Pamela Ventola, Kevin A Pelphrey, and James S Duncan. 2017. Identifying autism from resting-state fMRI using long short-term memory networks. In Machine Learning in Medical Imaging: 8th International Workshop, MLMI 2017, Held in Conjunction with MICCAI 2017, Q...

    2017

  4. [12]

    Alex Fornito, Andrew Zalesky, and Michael Breakspear. 2013. Graph analysis of the human connectome: promise, progress, and pitfalls. Neuroimage 80 (2013), 426–444

    2013

  5. [13]

    Xumeng Gong, Cheng Yang, and Chuan Shi. 2023. Ma-gcl: Model augmentation tricks for graph contrastive learning. In Proceedings of the AAAI Conference on Artificial Intelligence , Vol. 37. 4284–4292

    2023

  6. [14]

    Shih-Cheng Huang, Anuj Pareek, Malte Jensen, Matthew P Lungren, Serena Yeung, and Akshay S Chaudhari. 2023. Self-supervised learning for medical image classification: a systematic review and implementation guidelines. NPJ Digital Medicine 6, 1 (2023), 74

    2023

  7. [15]

    Weiyu Huang, Leah Goldsberry, Nicholas F Wymbs, Scott T Grafton, Danielle S Bassett, and Alejandro Ribeiro. 2016. Graph frequency analysis of brain signals. IEEE Journal of Selected Topics in Signal Processing 10, 7 (2016), 1189–1203

    2016

  8. [16]

    Zhe Huang, Ruijie Jiang, Shuchin Aeron, and Michael C Hughes. 2024. Systematic comparison of semi-supervised and self-supervised learning for medical image classification. In Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition . 22282–22293

    2024

  9. [17]

    R Matthew Hutchison, Thilo Womelsdorf, Elena A Allen, Peter A Bandettini, Vince D Calhoun, Maurizio Corbetta, Stefania Della Penna, Jeff H Duyn, Gary H Glover, Javier Gonzalez-Castillo, et al. 2013. Dynamic functional connectivity: promise, issues, and interpretations. Neuroim...

    2013

  10. [18]

    Susan L Hyman, Susan E Levy, Scott M Myers, Dennis Kuo, Susan Apkon, Timothy Brei, Lynn F Davidson, Beth Ellen Davis, Kathryn A Ellerbeck, Garey H Noritz, et al. 2020. Executive summary: identification, evaluation, and management of children with autism spectrum disorder. Pedi...

    2020

  11. [19]

    Alireza Talesh Jafadideh and Babak Mohammadzadeh Asl. 2022. Rest-fMRI based comparison study between autism spectrum disorder and typically control using graph frequency bands. Computers in Biology and Medicine 146 (2022), 105643

    2022

  12. [20]

    Smitha Karavallil Achuthan, Kelly L Coburn, Meagan E Beckerson, and Rajesh K Kana. 2023. Amplitude of low frequency fluctuations during resting state fMRI in autistic children. Autism Research 16, 1 (2023), 84–98

    2023

  13. [21]

    Pranjal Kumar, Piyush Rawat, and Siddhartha Chauhan. 2022. Contrastive self-supervised learning: review, progress, challenges and future research directions. International Journal of Multimedia Information Retrieval 11, 4 (2022), 461–488

    2022

  14. [22]

    Xiaoxiao Li, Yuan Zhou, Nicha Dvornek, Muhan Zhang, Siyuan Gao, Juntang Zhuang, Dustin Scheinost, Lawrence H Staib, Pamela Ventola, and James S Duncan. 2021. Braingnn: Interpretable brain graph neural network for fmri analysis. Medical Image Analysis 74 (2021), 102233

    2021

  15. [23]

    Mujie Liu, Qichao Dong, Chenze Wang, Xinrui Cheng, Falih Gozi Febrinanto, Azadeh Noori Hoshyar, and Feng Xia. 2024. Motif-induced Subgraph Generative Learning for Explainable Neurological Disorder Detection. InAustralasian Joint Conference on Artificial Intelligence. Springer, 376–389

    2024

  16. [24]

    Nian Liu, Xiao Wang, Deyu Bo, Chuan Shi, and Jian Pei. 2022. Revisiting graph contrastive learning from the perspective of graph spectrum. Advances in Neural Information Processing Systems 35 (2022), 2972–2983

    2022

  17. [25]

    Youfa Liu and Bo Du. 2024. Frequency domain-oriented complex graph neural networks for graph classification. IEEE Transactions on Neural Networks and Learning Systems (2024)

    2024

  18. [26]

    Yixin Liu, Ming Jin, Shirui Pan, Chuan Zhou, Yu Zheng, Feng Xia, and S Yu Philip. 2022. Graph self-supervised learning: A survey. IEEE transactions on knowledge and data engineering 35, 6 (2022), 5879–5900

    2022

  19. [27]

    Ilya Loshchilov and Frank Hutter. 2019. Decoupled weight decay regularization. ICLR (2019)

    2019

  20. [28]

    Xuexiong Luo, Guangwei Dong, Jia Wu, Amin Beheshti, Jian Yang, and Shan Xue. 2024. An Interpretable Brain Graph Contrastive Learning Framework for Brain Disorder Analysis. In Conference on Web Search and Data Mining (WSDM '24)

    2024

  21. [29]

    Xuexiong Luo, Jia Wu, Jian Yang, Hongyang Chen, Zhao Li, Hao Peng, and Chuan Zhou. 2024. Knowledge distillation guided interpretable brain subgraph neural networks for brain disorder exploration. IEEE Transactions on Neural Networks and Learning Systems 36, 2 (2024), 3559–3572

    2024

  22. [30]

    Xuexiong Luo, Jia Wu, Jian Yang, Shan Xue, Amin Beheshti, Quan Z Sheng, David McAlpine, Paul Sowman, Alexis Giral, and S Yu Philip. 2024. Graph neural networks for brain graph learning: a survey. In 33rd International Joint Conference on Artificial Intelligence, IJCAI 2024. In...

    2024

  23. [31]

    Sreevalsan S Menon and K Krishnamurthy. 2019. A comparison of static and dynamic functional connectivities for identifying subjects and biological sex using intrinsic individual brain connectivity. Scientific Reports 9, 1 (2019), 5729

    2019

  24. [32]

    Faustino Muetunda, Soumaya Sabry, M Luqman Jamil, Sebastião Pais, Gaël Dias, and João Cordeiro. 2025. AI-Assisted Diagnosing, Monitoring and Treatment of Mental Disorders: A Survey. ACM Transactions on Computing for Healthcare 5, 4 (2025), 1–24

    2025

  25. [33]

    Kevin Murphy and Michael D Fox. 2017. Towards a consensus regarding global signal regression for resting state functional connectivity MRI. Neuroimage 154 (2017), 169–173

    2017

  26. [34]

    Agnibh Pathak, Soham Bhattacharjee, Tulika Saha, and Sriparna Saha. 2024. Does Sentiment and Emotion affect Mental Health? A Multi-task Classification Framework for Comprehensive Understanding of Mental Health, Emotion, and Sentiment from Motivational Conversations. ACM Transa...

    2024

  27. [35]

    Ciyuan Peng, Jiayuan He, and Feng Xia. 2024. Learning on Multimodal Graphs: A Survey. ArXiv Preprint ArXiv:2402.05322 (2024). ACM Trans. Comput. Healthcare, Vol. 1, No. 1, Article . Publication date: September 2025. Data-Efficient Psychiatric Disorder Detection via Self-superv...

    2024 arXiv

  28. [36]

    Ciyuan Peng, Mujie Liu, Chenxuan Meng, Sha Xue, Kathleen Keogh, and Feng Xia. 2024. Stage-aware Brain Graph Learning for Alzheimer′ s Disease. MedRxiv (2024), 2024–04

    2024

  29. [37]

    Liang Peng, Nan Wang, Jie Xu, Xiaofeng Zhu, and Xiaoxiao Li. 2022. GATE: graph CCA for temporal SElf-supervised learning for label-efficient fMRI analysis. IEEE Transactions on Medical Imaging 42, 2 (2022), 391–402

    2022

  30. [38]

    Xihao Piao, Zheng Chen, Taichi Murayama, Yasuko Matsubara, and Yasushi Sakurai. 2024. Fredformer: Frequency debiased transformer for time series forecasting. In Proceedings of the 30th ACM SIGKDD Conference on Knowledge Discovery and Data Mining . 2400–2410

    2024

  31. [39]

    Darrel A Regier, Emily A Kuhl, and David J Kupfer. 2013. The DSM-5: Classification and criteria changes. World Psychiatry 12, 2 (2013), 92–98

    2013

  32. [40]

    Jing Ren, Feng Xia, Ivan Lee, Azadeh Noori Hoshyar, and Charu Aggarwal. 2023. Graph Learning for Anomaly Analytics: Algorithms, Applications, and Challenges. ACM Transactions on Intelligent Systems and Technology 14, 2, Article 28 (Feb 2023), 29 pages. https: //doi.org/10.1145/3570906

    2023 doi

  33. [41]

    Mikail Rubinov and Olaf Sporns. 2010. Complex network measures of brain connectivity: uses and interpretations. Neuroimage 52, 3 (2010), 1059–1069

    2010

  34. [42]

    Joana Cristo Santos, Hugo Tomás Pereira Alexandre, Miriam Seoane Santos, and Pedro Henriques Abreu. [n. d.]. The Role of Deep Learning in Medical Image Inpainting: A Systematic Review. ACM Transactions on Computing for Healthcare ([n. d.])

  35. [43]

    N Sartorius. 2017. Comorbidity of mental and physical disorders: A major challenge for medicine in the 21st century.European Psychiatry 41, S1 (2017), S9–S9

    2017

  36. [44]

    Shuntaro Sasai, Takahiko Koike, Sho K Sugawara, Yuki H Hamano, Motofumi Sumiya, Shuntaro Okazaki, Haruka K Takahashi, Gentaro Taga, and Norihiro Sadato. 2021. Frequency-specific task modulation of human brain functional networks: A fast fMRI study.NeuroImage 224 (2021), 117375

    2021

  37. [45]

    David I Shuman, Sunil K Narang, Pascal Frossard, Antonio Ortega, and Pierre Vandergheynst. 2013. The emerging field of signal processing on graphs: Extending high-dimensional data analysis to networks and other irregular domains. IEEE Signal Processing Magazine 30, 3 (2013), 83–98

    2013

  38. [46]

    Ke Sun, Ciyuan Peng, Shuo Yu, Zhuoyang Han, and Feng Xia. 2024. From EEG Data to Brain Networks: Graph Learning Based Brain Disease Diagnosis. IEEE Intelligent Systems (2024), 1–9. https://doi.org/10.1109/MIS.2024.3352972

    2024

  39. [47]

    Haoteng Tang, Guixiang Ma, Lei Guo, Xiyao Fu, Heng Huang, and Liang Zhan. 2022. Contrastive brain network learning via hierarchical signed graph pooling model. IEEE transactions on neural networks and learning systems (2022)

    2022

  40. [48]

    Shantanu Thakoor, Corentin Tallec, Mohammad Gheshlaghi Azar, Rémi Munos, Petar Veličković, and Michal Valko. 2021. Bootstrapped representation learning on graphs. In ICLR 2021 Workshop on Geometrical and Topological Representation Learning

    2021

  41. [49]

    Alphin J Thottupattu and Jayanthi Sivaswamy. 2024. A Method for Comparing Time Series by Untangling Time-Dependent and Independent Variations in Biological Processes. ACM Transactions on Computing for Healthcare 5, 4 (2024), 1–15

    2024

  42. [50]

    Xiaochuan Wang, Ying Chu, Qianqian Wang, Liang Cao, Lishan Qiao, Limei Zhang, and Mingxia Liu. 2023. Unsupervised contrastive graph learning for resting-state functional MRI analysis and brain disorder detection. Human Brain Mapping 44, 17 (2023), 5672–5692

    2023

  43. [51]

    Yu Wu, Ting Dang, Dimitris Spathis, Hong Jia, and Cecilia Mascolo. 2024. StatioCL: Contrastive Learning for Time Series via Non- Stationary and Temporal Contrast. In Proceedings of the 33rd ACM International Conference on Information and Knowledge Management . 2575–2584

    2024

  44. [52]

    Feng Xia, Ke Sun, Shuo Yu, Abdul Aziz, Liangtian Wan, Shirui Pan, and Huan Liu. 2021. Graph Learning: A Survey. IEEE Transactions on Artificial Intelligence 2, 2 (2021), 109–127. https://doi.org/10.1109/TAI.2021.3076021

    2021

  45. [53]

    Feng Xia, Lei Wang, Tao Tang, Xin Chen, Xiangjie Kong, Giles Oatley, and Irwin King. 2023. CenGCN: Centralized Convolutional Networks with Vertex Imbalance for Scale-Free Graphs. IEEE Transactions on Knowledge and Data Engineering 35, 5 (2023), 4555–4569. https://doi.org/10.11...

    2023

  46. [54]

    big data

    Mingrui Xia and Yong He. 2017. Functional connectomics from a “big data” perspective. Neuroimage 160 (2017), 152–167

    2017

  47. [55]

    Jiaxing Xu, Qingtian Bian, Xinhang Li, Aihu Zhang, Yiping Ke, Miao Qiao, Wei Zhang, Wei Khang Jeremy Sim, and Balázs Gulyás. 2024. Contrastive Graph Pooling for Explainable Classification of Brain Networks. IEEE Transactions on Medical Imaging (2024)

    2024

  48. [56]

    Ling Yang and Shenda Hong. 2022. Unsupervised time-series representation learning with iterative bilinear temporal-spectral fusion. In International Conference on Machine Learning . PMLR, 25038–25054

    2022

  49. [57]

    Zhenyu Yang, Yantao Li, and Gang Zhou. 2023. Ts-gan: Time-series gan for sensor-based health data augmentation. ACM Transactions on Computing for Healthcare 4, 2 (2023), 1–21

    2023

  50. [58]

    Kun Yi, Qi Zhang, Wei Fan, Hui He, Liang Hu, Pengyang Wang, Ning An, Longbing Cao, and Zhendong Niu. 2023. FourierGNN: Rethinking multivariate time series forecasting from a pure graph perspective. Advances in Neural Information Processing Systems 36 (2023)

    2023

  51. [59]

    Kun Yi, Qi Zhang, Wei Fan, Shoujin Wang, Pengyang Wang, Hui He, Ning An, Defu Lian, Longbing Cao, and Zhendong Niu. 2024. Frequency-domain MLPs are more effective learners in time series forecasting. Advances in Neural Information Processing Systems 36 (2024). ACM Trans. Compu...

    2024

  52. [60]

    Yuning You, Tianlong Chen, Yongduo Sui, Ting Chen, Zhangyang Wang, and Yang Shen. 2020. Graph contrastive learning with augmentations. Advances in Neural Information Processing Systems 33 (2020), 5812–5823

    2020

  53. [61]

    Hengrui Zhang, Qitian Wu, Junchi Yan, David Wipf, and Philip S Yu. 2021. From canonical correlation analysis to self-supervised graph neural networks. Advances in Neural Information Processing Systems 34 (2021), 76–89

    2021

  54. [62]

    Shengjie Zhang, Xiang Chen, Xin Shen, Bohan Ren, Ziqi Yu, Haibo Yang, Xi Jiang, Dinggang Shen, Yuan Zhou, and Xiao-Yong Zhang

  55. [63]

    Si Zhang, Hanghang Tong, Jiejun Xu, and Ross Maciejewski. 2019. Graph convolutional networks: a comprehensive review.Computational Social Networks 6, 1 (2019), 1–23

    2019

  56. [64]

    Xiang Zhang, Ziyuan Zhao, Theodoros Tsiligkaridis, and Marinka Zitnik. 2022. Self-supervised contrastive pre-training for time series via time-frequency consistency. Advances in Neural Information Processing Systems 35 (2022), 3988–4003

    2022

  57. [65]

    Yifei Zhang, Hao Zhu, Zixing Song, Piotr Koniusz, and Irwin King. 2023. Spectral feature augmentation for graph contrastive learning and beyond. In Proceedings of the AAAI Conference on Artificial Intelligence , Vol. 37. 11289–11297

    2023

  58. [66]

    Tian Zhou, Ziqing Ma, Qingsong Wen, Xue Wang, Liang Sun, and Rong Jin. 2022. Fedformer: Frequency enhanced decomposed transformer for long-term series forecasting. In International Conference on Machine Learning . PMLR, 27268–27286

    2022

  59. [67]

    Yanqiao Zhu, Yichen Xu, Feng Yu, Qiang Liu, Shu Wu, and Liang Wang. 2021. Graph contrastive learning with adaptive augmentation. In Proceedings of the Web Conference 2021 . 2069–2080

    2021

  60. [68]

    Xiaowei Zhuang, Zhengshi Yang, and Dietmar Cordes. 2020. A technical review of canonical correlation analysis for neuroscience applications. Human Brain Mapping 41, 13 (2020), 3807–3833

    2020

  61. [69]

    Yongcheng Zong, Qiankun Zuo, Michael Kwok-Po Ng, Baiying Lei, and Shuqiang Wang. 2024. A new brain network construction paradigm for brain disorder via diffusion-based graph contrastive learning.IEEE Transactions on Pattern Analysis and Machine Intelligence (2024). ACM Trans. ...

    2024

  62. [2023]

    Medical Image Analysis 90 (2023), 102932

    A-GCL: Adversarial graph contrastive learning for fMRI analysis to diagnose neurodevelopmental disorders. Medical Image Analysis 90 (2023), 102932

    2023

Pith tools

Reviewed August 5, 2026 · model on record in the stance chip above.