Pith. sign in

REVIEW 4 major objections 6 minor 82 references

A Privacy-Preserving Domain Adversarial Federated learning for multi-site brain functional connectivity analysis

T0 review · 4 major / 6 minor · reviewed 2026-08-09 · deepseek-v4-flash

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

Pith's one-line read A federated learning framework for fMRI, DAFed, splits each subject's brain connectivity into shared and site-specific features and, by training adversarially across hospitals without sharing patient data, claims the best average…

desk verdict Interesting combination of federated learning and domain adaptation for multi-site fMRI, but the core training step as written cannot be executed by local clients, so the reported gains are not reproducible. read the letter →

arxiv 2502.01885 v1 pith:NP2M5AZH submitted 2025-02-03 cs.LG cs.AIeess.IV

Yipu Zhang , Likai Wang , Kuan-Jui Su , Aiying Zhang , Hao Zhu , Xiaowen Liu , Hui Shen , Vince D. Calhoun
show 2 more authors
Yuping Wang Hongwen Deng
This is my paper · ORCID
classification cs.LGcs.AIeess.IV
keywords federatedlearningdomainadversarialtrainingfunctionalconnectivityfeaturedisentanglementcontrastiveresting-statefMRIautismspectrumdisordermildcognitiveimpairment
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

This paper claims that a federated learning framework can train a single classifier across multiple hospitals' fMRI datasets without moving patient data, and that this approach outperforms both single-site and earlier federated methods on autism spectrum disorder (ASD) and mild cognitive impairment (MCI) classification. On the ABIDE multi-site dataset, the labeled variant reaches an average accuracy of 0.764 and the unlabeled variant reaches 0.753, versus 0.709 for the strongest federated baseline; on ADNI-3 it reaches 0.718. The building blocks are feature disentanglement, which splits each subject's connectivity features into components shared across sites and components unique to each site, adversarial training that transfers knowledge from a labeled central site to unlabeled local sites, and contrastive learning that aligns the shared features. The paper also argues that its interpretability module recovers known brain regions and functional connections associated with ASD and MCI despite the noise added for privacy.

What carries the argument

The machinery has five pieces. A graph convolutional feature generator with Jumping Knowledge builds a spatiotemporal embedding from dynamic functional connectivity matrices. A feature disentangler splits that embedding into the domain-invariant component (features that stay consistent across sites) and the domain-specific component, with a Mutual Information Neural Estimator minimizing their overlap. Two discriminators then act adversarially: a classifier trained on the labeled source domain, and a domain identifier that tries to tell whether features came from the source or a target site, so the feature extractor must fool the identifier on the invariant component. A multi-head self-attention mechanism fuses the two disentangled components before classification, and a contrastive loss pulls each local site's invariant features toward the global model's previous invariant features while pushing away older local versions. Federated averaging with Gaussian noise on local updates completes each round, and the improved Score-CAM replaces gradient-based explanations with score-weighted channel masks so that the added noise does not corrupt the attribution maps.

What would settle it

A direct way to settle the central claim is to run Algorithm 1 under its stated privacy constraints and log whether the source-loss gradient used in line 7 can be computed at a local site without access to the central data. If local sites only receive the broadcast parameters and loss value, then the loss gradient with respect to local parameters can be computed for the target term but not for the source term; a working implementation must therefore transmit something beyond parameters, and checking what is transmitted (loss scalar, gradient surrogate, or data) determines whether the privacy claim holds.

Watch

Extended reading notes

Core claim

DAFed's central discovery is that the non-IID problem in multi-site fMRI can be handled by explicitly modeling each dataset's features as a mixture of a domain-invariant part (the same across scanners and institutions) and a domain-specific part (local peculiarities), and then training with an adversarial objective so the classifier relies on the invariant part while the local part is preserved. The authors report that this design lets one labeled site (the source) teach unlabeled sites (the targets) under federated constraints, and that on the ABIDE and ADNI-3 cohorts the method achieves the best average classification accuracy among the compared methods in both labeled and unlabeled variants. They further claim the approach is generalizable: the same framework works for ASD diagnosis on ABIDE and MCI detection on ADNI-3, and the improved Score-CAM identifies both common and site-specific brain regions that align with earlier findings on these disorders.

Load-bearing premise

The training loop assumes that each local site can compute the gradient of the central site's loss with respect to its own model parameters, even though the central site's data and the source loss never leave the central site, and the paper gives no mechanism for obtaining that gradient.

Editorial extensions

If this is right

  • Multi-site studies can pool fMRI knowledge without transmitting raw scans, since only noise-added model parameters move between sites.
  • Sites that have no labels can still obtain a working classifier, because the adversarial transfer trains the shared extractor on the labeled source and applies it to unlabeled targets.
  • Scanner and institution differences that break naive federated averaging can be absorbed by the disentanglement module, which is why the method reports higher and more stable accuracy than FedAvg, FedMoE, FedAlign, and FedCL.
  • The noise-tolerant Score-CAM module gives the collaborative model an interpretability mechanism, so the brain regions and connections it highlights can be inspected even with privacy noise present.

Reading between the lines

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

  • The paper does not specify how a local site obtains the gradient of the central site's loss that Algorithm 1 uses in its update step; a working deployment would likely need to broadcast the source loss value or a gradient surrogate, which would change the privacy profile.
  • Because the method always designates one labeled site as the source and treats others as targets, it is effectively a semi-supervised domain adaptation scheme wearing federated clothing, so its accuracy gains may track the quality of the adversarial alignment more than the federated averaging itself.
  • The reported brain regions (e.g., amygdala, inferior frontal gyrus, default-mode network nodes) are testable biomarkers: using the same pipeline with a different source site or a different atlas would show whether the identified regions are stable or an artifact of the chosen source domain.
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 / 6 minor

Summary. This paper proposes DAFed, a federated deep-learning framework for multi-site brain functional connectivity analysis. The method combines graph-convolutional feature extraction, feature disentanglement into domain-invariant and domain-specific components, multi-head attention integration, domain-adversarial training between a central labeled site and unlabeled local sites, contrastive learning on the domain-invariant representation, and an improved Score-CAM interpretation module. It is evaluated on ABIDE for ASD classification (four sites, 370 subjects) and on ADNI-3 for MCI classification (three scanners, 844 subjects). The authors report average accuracies of 0.764 (DAFed_L) and 0.753 (DAFed_U) on ABIDE, and 0.718 (DAFed_L) on ADNI-3, claiming superior performance over non-federated and federated baselines, and present ROI and functional-connectivity analyses.

Significance. If the described algorithm were executable and the evaluations were valid, the contribution would be practically important: it targets non-IID data in multi-site neuroimaging without sharing raw data, and it includes an interpretability module for biomarker discovery. The combination of federated learning, disentanglement, and adversarial domain adaptation is a sensible design, and the paper evaluates on two public cohorts with several baselines. The central algorithmic description, however, contains a gradient-computation error that prevents the procedure from running as written, and the evaluation protocol has outcome-dependent source selection and non-nested hyperparameter optimization. These issues are load-bearing for the headline accuracy claims, so the manuscript in its current form cannot support its conclusions.

major comments (4)
  1. [Section II-C, Algorithm 1 lines 4-7] The local update is not computable as written. The protocol broadcasts only the global parameters and the scalar loss Ltotal_source; a scalar loss carries no computational graph connecting source-domain samples to local parameters, so the derivative d(Ltotal_source)/dTheta_k is undefined at client k. Backpropagating through Ltotal_source would require either source data at every local site or an explicit per-parameter source-gradient message, and neither is described. Since line 7 of Algorithm 1 uses exactly this gradient, the algorithm cannot execute in any standard deep learning framework, and the accuracies in Tables 4 and 5 cannot be traced to the described federated procedure.
  2. [Results, ABIDE and ADNI-3 source selection] The source domain is selected on the basis of the same baseline accuracies used in the comparison: the text states that USM is designated as the source because non-federated methods achieved better results on USM, and GE is chosen for the same reason on ADNI-3. Because 5-fold cross-validation is performed on the same data, this is outcome-dependent selection and can inflate the reported improvement of DAFed; it undermines the claim that DAFed outperforms all baselines. A nested or pre-registered source-selection rule, or an analysis showing robustness across all possible source choices, is needed.
  3. [Parameters setup and Experiments] Hyperparameters lambda1, lambda2, the learning rates, and the privacy noise level alpha are tuned with grid search on the same datasets used for the reported 5-fold cross-validation, and the cross-validation is not nested. This makes the reported accuracies optimistic and prevents a clean comparison with the baselines, whose hyperparameter choices are not reported with the same detail. The authors should either use a held-out validation fold for tuning or report sensitivity analyses over the hyperparameter ranges.
  4. [Section II-D and Parameters setup] The privacy-preservation claim is not quantified. Adding Gaussian noise alpha*sigma to uploaded local models does not by itself constitute differential privacy; no privacy budget, composition analysis, or formal guarantee is provided, so the title's claim of being privacy-preserving is not supported. At minimum, the authors should state the precise privacy model and either provide a formal guarantee or soften the claim.
minor comments (6)
  1. [Introduction] The sentence beginning 'Although methods such as ComBat can correct batch effects [10, 77]' is duplicated in the introduction and should be corrected.
  2. [Section II-A, Eq. (2)] The notation 'mean(H) || max(H)' uses '||' for concatenation but the concatenation axis and the resulting dimension are not defined; please clarify.
  3. [Table 3] The dimension flow from the GCN outputs to the first MLP in the disentangler is not explained: the input dimension 480 is not derived from the concatenated embeddings defined in Eq. (3).
  4. [Algorithm 1] Line 9 averages Theta_k^t while line 7 updates Theta_k^{t+1}; the superscript/subscript indexing around the aggregation step is inconsistent and should be corrected.
  5. [Tables 4 and 5] The tables report only mean accuracy with standard deviation across folds; no paired significance tests or confidence intervals are given, so the claim that DAFed is 'best' is not statistically supported.
  6. [Reproducibility] The manuscript does not include a code or data availability statement, which impedes reproducibility of the described method and experiments.

Circularity Check

0 steps flagged · score 0.0 of 10

No significant circularity: DAFed's accuracy results are empirical benchmarks, not derivations that reduce to their inputs.

full rationale

The paper's central claim is an empirical accuracy comparison (Tables 4 and 5), and no equation or algorithm step defines the reported accuracy in terms of a fitted parameter. The DAFed objective (Eq. 11) combines classification, mutual-information, contrastive, and domain-discrimination losses with tuned weights; the reported numbers come from a 5-fold cross-validation evaluation and are not equal by construction to any input. The choice of USM/GE as the source site is made after observing baseline performance, which is a selection-bias concern rather than circularity: DAFed's score is not definitionally the baseline score. The paper's self-citations (e.g., [80] for dynamic functional connectivity and [6] as the source of the preprocessing protocol and comparison baselines) are background or baseline references and are not the load-bearing justification for the accuracy claim. The main correctness issue in Algorithm 1—local sites cannot backpropagate through a scalar Ltotal_source without source data—is an implementability gap, not a circularity, because it concerns whether the described procedure can run, not whether a result reduces to its input. No claim in the paper is self-definitional or a renamed fitted value, so the circularity score is 0.

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

The central claim relies on the disentanglement hypothesis and on an optimization protocol that is not fully specified: clients appear to need the source-domain loss gradient without source data. Key hyperparameters are tuned on the same datasets. No new physical entities are introduced.

free parameters (6)
  • λ1 (mutual information loss weight) = 1
    Selected by grid search on ABIDE and ADNI-3; no separate validation set is reported.
  • λ2 (contrastive loss weight) = 0.1
    Selected by grid search on the same datasets.
  • γ (schedule parameter in λp) = 10
    Hand-set to 10; controls ramp-up of domain adversarial weight.
  • α (privacy noise level) = 0.01
    Chosen by hand; noise added to local parameters has no formal privacy analysis.
  • learning rate = 0.0001 (ABIDE), 0.01 (ADNI)
    Set per dataset with different decay schedules.
  • contrastive queue length = 5
    Fixed at 5 to increase negative samples; no sensitivity analysis.
assumptions (4)
  • domain assumption Each dataset's feature embedding decomposes into a domain-invariant component and a domain-specific component.
    Section II-B states this as 'a key hypothesis is that each dataset consists of both a domain-invariant component (fdi) and a domain-specific component (fds).' The central algorithm depends on this decomposition being learnable and stable across sites.
  • ad hoc to paper Local sites can compute the gradient of the central source-domain loss with respect to their own parameters without accessing the source data.
    Algorithm 1 lines 5-7 and Section II-C steps 3-5 update local parameters using ∂(Ltotal_source + Ltotal_target_k)/∂Θk, but Ltotal_source is computed from the central site's private labeled data. No mechanism is provided for clients to obtain this gradient.
  • domain assumption Adding Gaussian noise to uploaded model parameters provides sufficient privacy protection.
    Section II-D step 5 and Parameters setup add εn ∼ N(0, ασ) with α=0.01; the paper provides no differential privacy bound or membership-inference analysis.
  • domain assumption The improved Score-CAM attribution reflects true biological relevance of ROIs.
    The interpretability section assumes model attention maps can be compared to known ASD/MCI literature without external validation of the attributions.

how reviews work

0 comments
Cite this review

Pith. "Pith review of A Privacy-Preserving Domain Adversarial Federated learning for multi-site brain functional connectivity analysis." pith.science (2026). https://pith.science/paper/NP2M5AZH

@misc{pith2026250201885,
  author       = {Pith},
  title        = {Pith review of: A Privacy-Preserving Domain Adversarial Federated learning for multi-site brain functional connectivity analysis},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/NP2M5AZH}},
  note         = {Machine review of arXiv:2502.01885}
}
read the original abstract

Resting-state functional magnetic resonance imaging (rs-fMRI) and its derived functional connectivity networks (FCNs) have become critical for understanding neurological disorders. However, collaborative analyses and the generalizability of models still face significant challenges due to privacy regulations and the non-IID (non-independent and identically distributed) property of multiple data sources. To mitigate these difficulties, we propose Domain Adversarial Federated Learning (DAFed), a novel federated deep learning framework specifically designed for non-IID fMRI data analysis in multi-site settings. DAFed addresses these challenges through feature disentanglement, decomposing the latent feature space into domain-invariant and domain-specific components, to ensure robust global learning while preserving local data specificity. Furthermore, adversarial training facilitates effective knowledge transfer between labeled and unlabeled datasets, while a contrastive learning module enhances the global representation of domain-invariant features. We evaluated DAFed on the diagnosis of ASD and further validated its generalizability in the classification of AD, demonstrating its superior classification accuracy compared to state-of-the-art methods. Additionally, an enhanced Score-CAM module identifies key brain regions and functional connectivity significantly associated with ASD and MCI, respectively, uncovering shared neurobiological patterns across sites. These findings highlight the potential of DAFed to advance multi-site collaborative research in neuroimaging while protecting data confidentiality.

Figures

Figures reproduced from arXiv: 2502.01885 by the authors.

Figure 2
Figure 2. Workflow of domain adversarial training. For the sample {xi, yi} in the source domain, the classification loss is given by ℒ𝐶𝐶 = 1 𝑛𝑛�𝑦𝑦𝑖𝑖𝑙𝑙𝑙𝑙𝑙𝑙 1 𝐶𝐶𝑠𝑠𝑠𝑠𝑠𝑠 �𝑓𝑓𝐹𝐹 𝑖𝑖 � + (1 − 𝑦𝑦𝑖𝑖)𝑙𝑙𝑙𝑙𝑙𝑙 1 1 − 𝐶𝐶𝑠𝑠𝑠𝑠𝑠𝑠 �𝑓𝑓𝐹𝐹 𝑖𝑖 � 𝑛𝑛 𝑖𝑖=1 (8) where and 𝑓𝑓𝐹𝐹 𝑖𝑖 is the integrated feature obtained from the multi-head self-attention mechanism. The classifier C consists of MLP parameterized by θC and outputs the predicted label. Note that … view at source ↗
Figure 5
Figure 5. The framework of the improved Score-CAM module. The process begins by extracting the activation map from the output of the l-th GCN layer. Each channel of the activation map corresponds to the same feature across different ROIs and is then broadcasted channel-by-channel. These broadcasted activation maps are used as masks to the input. For each masked input, the forward-pass score corresponding to the target class i… view at source ↗

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

82 extracted references · 72 canonical work pages

  1. [1]

    What does fMRI tell us about neuronal activity? Nature reviews neuroscience 3, 142–151 (2002)

    Heeger, David J., and David Ress . What does fMRI tell us about neuronal activity? Nature reviews neuroscience 3, 142–151 (2002)

    2002

  2. [2]

    Stam, C. J. Modern network science of neurological disorders. Nature Reviews Neuroscience 15, 683–695 (2014)

    2014

  3. [3]

    Hulshoff Pol

    Van Den Heuvel, Martijn P., and Hilleke E. Hulshoff Pol. Exploring the brain network: a review on resting- state fMRI functional connectivity. European neuropsychopharmacology 20, 519–534 (2010)

    work page 2010

  4. [4]

    State-space model with deep learning for functional dynamics estimation in resting-state fMRI

    Suk, Heung-Il, et al. State-space model with deep learning for functional dynamics estimation in resting-state fMRI. NeuroImage 129, 292–307 (2016)

    work page 2016

  5. [5]

    Deep Learning in Medical Image Analysis

    Shen, Dinggang, Guorong Wu, and Heung-Il Suk. Deep Learning in Medical Image Analysis. Annual Review of Biomedical Engineering 19, 221–248 (2017)

    work page 2017

  6. [6]

    Li, X. et al. Multi-site fMRI analysis using privacy-preserving federated learning and domain adaptation: ABIDE results. Medical image analysis 65, 101765 (2020)

    work page 2020

  7. [7]

    Thompson

    Shen, Li, and Paul M. Thompson. Brain imaging genomics: integrated analysis and machine learning. Proceedings of the IEEE 108, 125–162 (2019)

    work page 2019

  8. [8]

    Harmonization of cortical thickness measurements across scanners and sites

    Fortin, Jean-Philippe, et al. Harmonization of cortical thickness measurements across scanners and sites. Neuroimage 167, 104–120 (2018)

    work page 2018

Show all 82 references
  1. [9]

    FedCL: Federated contrastive learning for multi- center medical image classification

    Liu, Zhenbing, et al . FedCL: Federated contrastive learning for multi- center medical image classification. Pattern Recognition 143, 109739 (2023)

    2023

  2. [10]

    Evan, Cheng Li, and Ariel Rabinovic

    Johnson, W. Evan, Cheng Li, and Ariel Rabinovic. Adjusting batch effects in microarray expression data using empirical Bayes methods. Biostatistics 8, 118–127 (2007)

    2007

  3. [11]

    Federated Learning for Smart Healthcare: A Survey

    Nguyen, Dinh C., et al . Federated Learning for Smart Healthcare: A Survey. ACM Comput. Surv. 55, 1–37 (2023)

    2023

  4. [12]

    Federated optimization in heterogeneous networks

    Li, Tian, et al. Federated optimization in heterogeneous networks. Proceedings of Machine learning and systems 2, 429–450 (2020)

    2020

  5. [13]

    Federated learning with matched averaging

    Wang, Hongyi, et al . Federated learning with matched averaging. arXiv preprint arXiv: 2002.06440 (2020)

    2020 arXiv

  6. [14]

    Federated learning with unbiased gradient aggregation and controllable meta updating

    Yao, Xin, et al. Federated learning with unbiased gradient aggregation and controllable meta updating. arXiv preprint arXiv: 1910.08234 (2019)

    2019 arXiv

  7. [15]

    Fedbn: Federated learning on non-iid features via local batch normalization

    Li, Xiaoxiao, et al. Fedbn: Federated learning on non-iid features via local batch normalization. arXiv preprint arXiv: 2102.07623 (2021)

    2021 arXiv

  8. [16]

    Communication -efficient learning of deep networks from decentralized data

    McMahan, Brendan, et al . Communication -efficient learning of deep networks from decentralized data. Artificial intelligence and statistics, 1273-1282 (PMLR 2017)

    2017

  9. [17]

    Federated learning on non-IID data: A survey

    Zhu, Hangyu, et al. Federated learning on non-IID data: A survey. Neurocomputing 465, 371– 390 (2021)

    2021

  10. [18]

    A survey of multi -source domain adaptation

    Sun, Shiliang, Honglei Shi, and Yuanbin Wu. A survey of multi -source domain adaptation. Information Fusion 24, 84–92 (2015)

    2015

  11. [19]

    Domain agnostic learning with disentangled representations

    Peng, Xingchao, et al . Domain agnostic learning with disentangled representations. in International conference on machine learning 5102–5112 (PMLR, 2019)

    2019

  12. [20]

    The autism brain imaging data exchange: towards a large -scale evaluation of the intrinsic brain architecture in autism

    Di Martino, Adriana, et al. The autism brain imaging data exchange: towards a large -scale evaluation of the intrinsic brain architecture in autism. Molecular psychiatry 19, 659–667 (2014)

    2014

  13. [21]

    The Alzheimer’s disease neuroimaging initiative (ADNI): MRI methods

    Jack Jr, Clifford R., et al . The Alzheimer’s disease neuroimaging initiative (ADNI): MRI methods. Magnetic Resonance Imaging 27, 685–691 (2008)

    2008

  14. [22]

    Representation learning on graphs with jumping knowledge networks

    Xu, Keyulu, et al . Representation learning on graphs with jumping knowledge networks. in International conference on machine learning, 5453–5462 (PMLR, 2018)

    2018

  15. [23]

    Mutual Information Neural Estimation

    Belghazi, Mohamed Ishmael, et al. Mutual Information Neural Estimation. in Proceedings of the 35th International Conference on Machine Learning 531–540 (PMLR, 2018)

    2018

  16. [24]

    Vaswani, A. et al. Attention is All you Need. in Advances in Neural Information Processing Systems 30 (2017)

    2017

  17. [25]

    Learning deep features for discriminative localization

    Zhou, Bolei, et al . Learning deep features for discriminative localization. in Proceedings of the IEEE conference on computer vision and pattern recognition, 2921–2929 (2016)

    2016

  18. [26]

    Score -CAM: Score -weighted visual explanations for convolutional neural networks

    Wang, Haofan, et al . Score -CAM: Score -weighted visual explanations for convolutional neural networks. in Proceedings of the IEEE/CVF conference on computer vision and pattern recognition workshops 24–25 (2020)

    2020

  19. [27]

    The neuro bureau preprocessing initiative: open sharing of preprocessed neuroimaging data and derivatives

    Craddock, Cameron, et al . The neuro bureau preprocessing initiative: open sharing of preprocessed neuroimaging data and derivatives. Frontiers in Neuroinformatics 7, 5 (2013)

    2013

  20. [28]

    Yang, X. et al. Functional Connectivity Network Fusion with Dynamic Thresholding for MCI Diagnosis. in Machine Learning in Medical Imaging (eds. Wang, L., Adeli, E., Wang, Q., Shi, Y . & Suk, H.-I.) vol. 10019 246–253 (Springer International Publishing, Cham, 2016)

    2016

  21. [29]

    & Suzuki, T

    Oono, K. & Suzuki, T. Graph Neural Networks Exponentially Lose Expressive Power for Node Classification. Preprint at https://doi.org/10.48550/arXiv.1905.10947 (2021)

    2021 doi arXiv

  22. [30]

    & Wang, Y

    Cai, C. & Wang, Y . A Note on Over -Smoothing for Graph Neural Networks. Preprint at https://doi.org/10.48550/arXiv.2006.13318 (2020)

    2020 doi arXiv

  23. [31]

    pipeline

    Yan, C. & Zang, Y . DPARSF: a MATLAB toolbox for" pipeline" data analysis of resting-state fMRI. Frontiers in systems neuroscience 4, 1377 (2010)

    2010

  24. [32]

    Ganin, Y . et al. Domain-adversarial training of neural networks. Journal of machine learning research 17, 1–35 (2016)

    2016

  25. [33]

    Kalra, D. S. & Barkeshli, M. Why Warmup the Learning Rate? Underlying Mechanisms and Improvements. Preprint at https://doi.org/10.48550/arXiv.2406.09405 (2024)

    2024 doi arXiv

  26. [34]

    & Jordan, M

    You, K., Long, M., Wang, J. & Jordan, M. I. How Does Learning Rate Decay Help Modern Neural Networks? Preprint at https://doi.org/10.48550/arXiv.1908.01878 (2019)

    2019 doi arXiv

  27. [35]

    Kipf, T. N. & Welling, M. Semi -Supervised Classification with Graph Convolutional Networks. Preprint at https://doi.org/10.48550/arXiv.1609.02907 (2017)

    2017 doi arXiv

  28. [36]

    & Hinton, G

    Van der Maaten, L. & Hinton, G. Visualizing data using t-SNE. Journal of machine learning research 9, (2008)

    2008

  29. [37]

    Selvaraju, R. R. et al. Grad-cam: Visual explanations from deep networks via gradient-based localization. in Proceedings of the IEEE international conference on computer vision 618– 626 (2017)

    2017

  30. [38]

    & Balasubramanian, V

    Chattopadhay, A., Sarkar, A., Howlader, P. & Balasubramanian, V . N. Grad -cam++: Generalized gradient -based visual explanations for deep convolutional networks. in 2018 IEEE winter conference on applications of computer vision (WACV) 839–847 (IEEE, 2018)

    2018

  31. [39]

    Thomas Yeo, B. T. et al. The organization of the human cerebral cortex estimated by intrinsic functional connectivity. Journal of Neurophysiology 106, 1125–1165 (2011)

    2011

  32. [40]

    Xia, M., Wang, J. & He, Y . BrainNet Viewer: a network visualization tool for human brain connectomics. PloS one 8, e68910 (2013)

    2013

  33. [41]

    Autism spectrum disorder characterization in children by capturing local-regional brain changes in MRI

    Alvarez-Jimenez, Charlems, et al . Autism spectrum disorder characterization in children by capturing local-regional brain changes in MRI. Medical Physics 47, 119–131 (2020)

    2020

  34. [42]

    EAG-RS: A Novel Explainability-guided ROI-Selection Framework for ASD Diagnosis via Inter-regional Relation Learning

    Jung, Wonsik, et al. EAG-RS: A Novel Explainability-guided ROI-Selection Framework for ASD Diagnosis via Inter-regional Relation Learning. IEEE Transactions on Medical Imaging (2023)

    2023

  35. [43]

    The functional neural architecture of dysfunctional reward processing in autism

    Janouschek, Hildegard, et al . The functional neural architecture of dysfunctional reward processing in autism. NeuroImage: Clinical 31, 102700 (2021)

    2021

  36. [44]

    Amygdala V olumes in Autism Spectrum Disorders: Meta-analysis of Magnetic Resonance Imaging Studies

    Kovacevic, Marija, et al. Amygdala V olumes in Autism Spectrum Disorders: Meta-analysis of Magnetic Resonance Imaging Studies. Rev J Autism Dev Disord 10, 169–183 (2023)

    2023

  37. [45]

    Nicholas, et al

    Top Jr, D. Nicholas, et al. Atypical amygdala response to fear conditioning in autism spectrum disorder. Biological Psychiatry: Cognitive Neuroscience and Neuroimaging 1, 308–315 (2016)

    2016

  38. [46]

    Individual differences in intrinsic brain networks predict symptom severity in autism spectrum disorders

    Pua, Emmanuel Peng Kiat, et al . Individual differences in intrinsic brain networks predict symptom severity in autism spectrum disorders. Cerebral Cortex 31, 681–693 (2021)

    2021

  39. [47]

    Improving social cognition following theta burst stimulation over the right inferior frontal gyrus in autism spectrum: an 8-week double-blind sham-controlled trial

    Ni, Hsing-Chang, et al. Improving social cognition following theta burst stimulation over the right inferior frontal gyrus in autism spectrum: an 8-week double-blind sham-controlled trial. Psychological Medicine 54, 3261–3272 (2024)

    2024

  40. [48]

    Abnormal activation of the social brain network in children with autism spectrum disorder: an FMRI study

    Kim, Sun-Young, et al . Abnormal activation of the social brain network in children with autism spectrum disorder: an FMRI study. Psychiatry investigation 12, 37 (2015)

    2015

  41. [49]

    Subcortical brain morphometry differences between adults with autism spectrum disorder and schizophrenia

    Weerasekera, Akila, et al . Subcortical brain morphometry differences between adults with autism spectrum disorder and schizophrenia. Brain Sciences 12, 439 (2022)

    2022

  42. [50]

    Specific Functional Connectivity Patterns of Middle Temporal Gyrus Subregions in Children and Adults with Autism Spectrum Disorder

    Xu, Jinping, et al . Specific Functional Connectivity Patterns of Middle Temporal Gyrus Subregions in Children and Adults with Autism Spectrum Disorder. Autism Research 13, 410– 422 (2020)

    2020

  43. [51]

    Do 2-year changes in superior frontal gyrus and global brain atrophy affect cognition? Alzheimer's & Dementia: Diagnosis, Assessment & Disease Monitoring 10, 706–716 (2018)

    Valdés Hernández, Maria del C., et al. Do 2-year changes in superior frontal gyrus and global brain atrophy affect cognition? Alzheimer's & Dementia: Diagnosis, Assessment & Disease Monitoring 10, 706–716 (2018)

    2018

  44. [52]

    Lateral temporal hyper -activation as a novel biomarker of mild cognitive impairment

    Faraco, Carlos C., et al . Lateral temporal hyper -activation as a novel biomarker of mild cognitive impairment. Neuropsychologia 51, 2281–2293 (2013)

    2013

  45. [53]

    Integrated cerebellar radiomic-network model for predicting mild cognitive impairment in Alzheimer's disease

    Chen, Yini, et al. Integrated cerebellar radiomic-network model for predicting mild cognitive impairment in Alzheimer's disease. Alzheimer's & Dementia (2024)

    2024

  46. [54]

    ASMFS: Adaptive -similarity-based multi -modality feature selection for classification of Alzheimer’s disease

    Shi, Yuang, et al . ASMFS: Adaptive -similarity-based multi -modality feature selection for classification of Alzheimer’s disease. Pattern Recognition 126, 108566 (2022)

    2022

  47. [55]

    Anatomical and functional deficits in patients with amnestic mild cognitive impairment

    Han, Ying, et al. Anatomical and functional deficits in patients with amnestic mild cognitive impairment. PloS one 7, e28664 (2012)

    2012

  48. [56]

    Enriched white matter connectivity networks for accurate identification of MCI patients

    Wee, Chong- Yaw, et al . Enriched white matter connectivity networks for accurate identification of MCI patients. Neuroimage 54, 1812–1822 (2011)

    2011

  49. [57]

    Characterization of resting state activity in MCI individuals

    Esposito, Roberto, et al. Characterization of resting state activity in MCI individuals. PeerJ 1, e135 (2013)

    2013

  50. [58]

    Structural and functional abnormalities of olfactory -related regions in subjective cognitive decline, mild cognitive impairment, and Alzheimer’s disease

    Chen, Ben, et al. Structural and functional abnormalities of olfactory -related regions in subjective cognitive decline, mild cognitive impairment, and Alzheimer’s disease. International Journal of Neuropsychopharmacology 25, 361–374 (2022)

    2022

  51. [59]

    MRI morphometry of the anterior and posterior cerebellar vermis and its relationship to sensorimotor and cognitive functions in children

    Hodgdon, Elizabeth A., et al . MRI morphometry of the anterior and posterior cerebellar vermis and its relationship to sensorimotor and cognitive functions in children. Developmental Cognitive Neuroscience 67, 101385 (2024)

    2024

  52. [60]

    Absence of associations with prefrontal cortex and cerebellum may link to early language and social deficits in preschool children with ASD

    Xin, Jing, et al. Absence of associations with prefrontal cortex and cerebellum may link to early language and social deficits in preschool children with ASD. Frontiers in Psychiatry 14, 1144993 (2023)

    2023

  53. [61]

    Visual perception in autism spectrum disorder: a review of neuroimaging studies

    Chung, Seungwon, and Jung- Woo Son. Visual perception in autism spectrum disorder: a review of neuroimaging studies. Journal of the Korean Academy of Child and Adolescent Psychiatry 31, 105 (2020)

    2020

  54. [62]

    Age-related differences in the intrinsic connectivity of the hippocampal and ventral temporal lobe in autistic individuals

    Chen, Lang, et al. Age-related differences in the intrinsic connectivity of the hippocampal and ventral temporal lobe in autistic individuals. Frontiers in Human Neuroscience 18, 1394706 (2024)

    2024

  55. [63]

    Reduced cortical thickness of the paracentral lobule in at-risk mental state individuals with poor 1- year functional outcomes

    Sasabayashi, Daiki, et al. Reduced cortical thickness of the paracentral lobule in at-risk mental state individuals with poor 1- year functional outcomes. Translational psychiatry 11, 396 (2021)

    2021

  56. [64]

    Supplementary motor area structure and function: review and hypotheses

    Goldberg, Gary. Supplementary motor area structure and function: review and hypotheses. Behavioral and brain Sciences 8, 567–588 (1985)

    1985

  57. [65]

    Anatomy and white matter connections of the lingual gyrus and cuneus

    Palejwala, Ali H., et al. Anatomy and white matter connections of the lingual gyrus and cuneus. World neurosurgery 151, e426–e437 (2021)

    2021

  58. [66]

    Gyri of the human parietal lobe: volumes, spatial extents, automatic labelling, and probabilistic atlases

    Goldberg, Gary.Wild, Heather M., et al. Gyri of the human parietal lobe: volumes, spatial extents, automatic labelling, and probabilistic atlases. PloS one 12, e0180866 (2017)

    2017

  59. [67]

    20 years of the default mode network: A review and synthesis

    Menon, Vinod. 20 years of the default mode network: A review and synthesis. Neuron 111, 2469–2487 (2023)

    2023

  60. [68]

    The Organization and Operation of Inferior Temporal Cortex

    Conway, Bevil R. The Organization and Operation of Inferior Temporal Cortex. Annu. Rev. Vis. Sci. 4, 381–402 (2018)

    2018

  61. [69]

    Anterior cingulate cortex activation during attentional control as a transdiagnostic marker of psychotherapy response: A randomized clinical trial

    Feurer, Cope, et al . Anterior cingulate cortex activation during attentional control as a transdiagnostic marker of psychotherapy response: A randomized clinical trial. Neuropsychopharmacology 47, 1350–1357 (2022)

    2022

  62. [70]

    Parietal cortex matters in Alzheimer’s disease: an overview of structural, functional and metabolic findings

    Jacobs, Heidi IL, et al . Parietal cortex matters in Alzheimer’s disease: an overview of structural, functional and metabolic findings. Neuroscience & Biobehavioral Reviews 36, 297–309 (2012)

    2012

  63. [71]

    Fractional anisotropy changes in parahippocampal cingulum due to Alzheimer’s disease

    Dalboni da Rocha, Josué Luiz, et al . Fractional anisotropy changes in parahippocampal cingulum due to Alzheimer’s disease. Scientific reports 10, 2660 (2020)

    2020

  64. [72]

    Atypical cross talk between mentalizing and mirror neuron networks in autism spectrum disorder

    Fishman, Inna, et al. Atypical cross talk between mentalizing and mirror neuron networks in autism spectrum disorder. JAMA psychiatry 71, 751–760 (2014)

    2014

  65. [73]

    Characterization of Autism Spectrum Disorder across the Age Span by Intrinsic Network Patterns

    Morgan, Benjamin R., et al . Characterization of Autism Spectrum Disorder across the Age Span by Intrinsic Network Patterns. Brain Topogr 32, 461–471 (2019)

    2019

  66. [74]

    Dysmaturation of the default mode network in autism

    Washington, Stuart D., et al. Dysmaturation of the default mode network in autism. Human Brain Mapping 35, 1284–1296 (2014)

    2014

  67. [75]

    Imaging brain amyloid in Alzheimer ’s disease with Pittsburgh Compound-B

    Klunk, William E., et al . Imaging brain amyloid in Alzheimer ’s disease with Pittsburgh Compound-B. Annals of Neurology 55, 306–319 (2004)

    2004

  68. [76]

    Default mode network functional connectivity in early and late mild cognitive impairment: results from the Alzheimer’s disease neuroimaging initiative

    Lee, Eek-Sung, et al. Default mode network functional connectivity in early and late mild cognitive impairment: results from the Alzheimer’s disease neuroimaging initiative. Alzheimer Disease & Associated Disorders 30, 289–296 (2016)

    2016

  69. [77]

    A decentralized ComBat algorithm and applications to functional network connectivity

    Bostami, Biozid, et al . A decentralized ComBat algorithm and applications to functional network connectivity. Frontiers in Neurology 13, 826734 (2022)

    2022

  70. [78]

    Federated analysis of neuroimaging data: a review of the field

    Rootes-Murdy, Kelly, et al. Federated analysis of neuroimaging data: a review of the field. Neuroinformatics 20, 377–390 (2022)

    2022

  71. [79]

    Enhancing collaborative neuroimaging research: introducing COINSTAC Vaults for federated analysis and reproducibility

    Martin, Dylan, et al . Enhancing collaborative neuroimaging research: introducing COINSTAC Vaults for federated analysis and reproducibility. Frontiers in Neuroinformatics 17, 1207721 (2023)

    2023

  72. [80]

    The chronnectome: time-varying connectivity networks as the next frontier in fMRI data discovery

    Calhoun, Vince D., et al. The chronnectome: time-varying connectivity networks as the next frontier in fMRI data discovery. Neuron 84, 2: 262-274 (2014)

    2014

  73. [81]

    Sharing privacy-sensitive access to neuroimaging and genetics data: a review and preliminary validation

    Sarwate, Anand D., et al. Sharing privacy-sensitive access to neuroimaging and genetics data: a review and preliminary validation. Frontiers in neuroinformatics 8, 35 (2014)

    2014

  74. [82]

    Privacy‐preserving quality control of neuroimaging datasets in federated environments

    Saha, Debbrata K., et al . Privacy‐preserving quality control of neuroimaging datasets in federated environments. Human Brain Mapping 43, 7: 2289-2310 (2022). Acknowledgements This research was supported in part by the National Center for Advancing Translational Research of NI...

    2022

Pith tools

Reviewed August 9, 2026 · model on record in the stance chip above.