Pith. sign in

REVIEW 3 major objections 4 minor 104 references

CyberNeuro: A Privacy-Preserving Agentic Workbench for Cohort-Scale Neuroimage and Clinical Data Analysis

T0 review · 3 major / 4 minor · reviewed 2026-08-03 · deepseek-v4-flash

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

Pith's one-line read A local 7B model, tuning only ~1% of its weights, drives the neuroimaging workflows that currently need cloud agents — held-out accuracy jumps from 40% to 69%, and data never leaves the machine.

desk verdict Serious systems paper with an honest results section, but the 40→69 held-out accuracy claim needs a clearer decontamination statement before it can be trusted. read the letter →

arxiv 2607.28841 v1 pith:FTCGOWUD submitted 2026-07-30 cs.MA cs.SE

Ran Ren , Junhong Tong , Yunxi Kong , Yiyao Chen , Yucheng Li , Kunhao Zhou , Shaoqi Wang , Yuxiang Tao
show 5 more authors
Shuheng Cao Zhihao Fan Marissa DiPiero Tingting Dan Guorong Wu
This is my paper · ORCID
classification cs.MAcs.SE
keywords agenticneuroimaginglocallanguagemodelprivacy-preservingcomputingDICOM-to-BIDSconversionsafeabstentionQLoRAadaptationmulti-agentorchestrationbrainimagingdatastructure(BIDS)
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's central claim is that clinical neuroimaging workflows — converting raw DICOM scans into the standard BIDS layout, running processing pipelines, and quality-checking outputs — can be automated end to end by an agentic system whose entire language model runs on the researcher's own workstation, with no data leaving the machine. Its specialist 7B model, WandaMind, lifts accuracy on a held-out 62-case suite from 40.3% to 69.4%, and the paper shows the gain is not better tool-following but safe abstention: false tool calls drop from 100% to 0%. At the system level, ten heterogeneous cohorts are converted to BIDS-validator-clean datasets with zero human interventions — versus 24 for a cloud-based comparison agent — and about 89% fewer tokens per converted volume. Under process-level network denial, all tested tasks complete with zero sent bytes, supporting the local-first privacy claim. The consequence, if the results hold: laboratories bound by data-use agreements and without cloud budgets can bring agentic automation to protected imaging data on commodity hardware.

What carries the argument

WandaMind's decision boundary, enforced by a deterministic validation layer, carries the argument. WandaMind is a 7B base adapted with response-only QLoRA (parameter-efficient fine-tuning on quantized weights; 1.06% of parameters trainable) in three rounds — supervised initialization, act-only augmentation, adversarial hardening — on a corpus decontaminated against QC-62 by token-level Jaccard similarity. The orchestration layer (Planner, Validator, Dispatcher, Reporter) keeps the LLM routing-only: nothing executes until validation passes, and a run succeeds only when artifact checks confirm a genuine derivative — without this, all four malformed cases exited 0 and were reported complete. Ov

What would settle it

Release the full training corpus — all three rounds, including the 21 hardening counterexamples and the act-only augmentations — and run both token-level and semantic (embedding-similarity) decontamination against the 62 QC-62 prompts. Any semantic near-duplicate beyond the stated Jaccard threshold would inflate the 69.4% figure, while a clean audit would vindicate it. Independently, a prospective replication on unseen scanners and sites, demanding the same zero-intervention, BIDS-validator-clean outcome, would test whether the architecture rather than the specific test set carries the result.

Watch

Extended reading notes

Core claim

A small local language model can orchestrate real neuroimaging analysis if orchestration is 'LLM-routed but tool-deterministic': the model proposes, deterministic layers dispose. WandaMind, a 7B base specialized via response-only QLoRA on ~700 decontaminated trajectories plus adversarial hardening, scores 69.4% on held-out QC-62 versus 40.3% (base) and 48.4% (27B); the entire margin is decision quality (abstain recall 0→100%, false actions 100→0%), tool selection unchanged. End to end, ten cohorts convert to validator-clean BIDS datasets in one request each, with 0 interventions and 1,127.5 tokens per volume versus 10,646.2 for the comparison agent; network denial shows zero sent bytes. The

Load-bearing premise

The load-bearing premise is that the held-out QC-62 cases were never used, even indirectly, to shape WandaMind's behavior: the paper's decontamination (§3.7.1) removes training candidates only by token-level Jaccard similarity ≥ 0.55, which cannot catch semantic near-duplicates, so if any adversarial-hardening counterexample or act-only augmentation was derived from a QC-62 case, the 69.4% accuracy is inflated.

Editorial extensions

If this is right

  • Labs restricted by IRB and data-use agreements can run agentic automation without transmitting protected imaging data: the default configuration is fully local, and process-level network denial showed zero sent bytes for all tested tasks.
  • DICOM-to-BIDS conversion becomes schedulable unattended: zero human interventions across ten cohorts, versus 24 for the comparison agent — the paper states this as a capability boundary, not a margin.
  • Token economics favor local orchestration: about one-ninth the tokens per NIfTI volume with the local model, and a roughly 29x gap on total tokens per cohort because the local system triages away non-anatomical series instead of converting everything.
  • The success criterion is what makes unattended execution trustworthy: a run counts as successful only if the expected scientific derivative exists and passes output checks — the failure-injection test shows that without artifact validation, all four malformed cases were reported complete with exit code 0.
  • In bounded tool-use domains, a specialized 7B model can outperform a 27B model because the larger model acts on half of the cases it should refuse; the binding constraint is calibrated abstention, not raw capability.

Reading between the lines

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

  • Scale may be the wrong lever for bounded tool orchestration: the evidence puts decision calibration and deterministic enforcement ahead of model size, implying that reliable agents in constrained tool domains might be bought with small specialized models plus validators rather than with larger base models.
  • The routing-only contract — LLM proposes, validator disposes, success requires an inspectable artifact — is data-agnostic, so a direct extension would port it to other protected biomedical pipelines, such as pathology-slide triage or genomic variant curation, wherever a deterministic checker can verify the derivative.
  • The paper's stated long-context weakness suggests a hybrid it leaves implicit: keep the local 7B as the default router and escalate genuinely hard planning to a stronger model inside the same validation envelope, using the measured token savings as the budget for such escalation.
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

3 major / 4 minor

Summary. The paper introduces CyberNeuro, a Windows desktop agentic workbench for neuroimaging and clinical data analysis that keeps all model inference, tool execution, and data within a local privacy boundary. Four agents (Planner, Validator, Dispatcher, Reporter) mediate between a natural-language interface and a pinned skill library via a Model Context Protocol bridge. The core technical contribution is WandaMind, a QLoRA-adapted Qwen2.5-7B model trained with response-only supervision, token-level decontamination against a held-out QC-62 set, and three rounds of iterative failure mining and adversarial hardening. The paper reports that WandaMind improves held-out tool-use accuracy from 40.3% to 69.4%, reduces DICOM-to-BIDS human interventions from 24 to 0 across ten cohorts, and consumes about 89% fewer tokens per NIfTI than NeuroClaw. It also reports post-install network denial tests and claims zero outbound traffic in the disconnected condition. The authors are explicit that the 27B-comparator advantage is not significant and that per-NIfTI normalization favors NeuroClaw because of different output counts.

Significance. If the claims hold, the paper is a useful systems contribution: it demonstrates that a small locally deployed model can be hardened to route and refuse within a bounded neuroimaging tool space, and that deterministic validation gates can make unattended execution safe enough for cohort-scale conversion. The strengths are the explicit separation of LLM planning from deterministic contract, policy, and artifact validation; the provenance receipts; the reproduction of the 10-cohort comparison under two backends; and the candid treatment of limitations (host-wide netstat not process-attributed, per-NIfTI normalization favoring the baseline, adaptation being decision hardening rather than general capability gain). However, the central 69.4% held-out claim is not yet credible because the provenance of the Round 2/3 hardening examples relative to QC-62 is not disclosed, and the abstract overstates the evaluation as being on the public NeuroBench suite. The token-efficiency headline is also stated inconsistently in the abstract, though the body is more careful.

major comments (3)
  1. [§3.7.3, §4.1 (Table 3)] Section 3.7.3 describes three training rounds, including Round 2 'ACT-Only Augmentation' motivated by 'four diagnostic cases' and Round 3 'Adversarial Hardening' with '+21 Hardening Counterexamples', but it never states that these examples were passed through the Jaccard decontamination against QC-62 described in §3.7.1 (which is phrased only for the 'core pool of 722 training trajectories'). The 29.1-point held-out gain in Table 3 is concentrated in exactly the categories those rounds targeted (Quality Judgment 0/11→7/11, Command Construction 0/6→4/6, Escalation 0/4→4/4, Fig. 2d). If any diagnostic or hardening example was derived from QC-62 cases, even as a paraphrase below the Jaccard threshold, the claimed held-out accuracy is inflated. The paper provides no evidence that QC-62 remained unseen during iterative development. Please disclose the provenance of the 4 diagnostic cases and
  2. [Abstract, §4.1, §8] The Abstract and Conclusion state that the held-out accuracy improvement (40%→69%) is demonstrated 'on the public NeuroBench suite,' but the model evaluation in §4.1 is described only on 'QC-62, a held-out collection of 62 neuroimaging tool-use and quality-control cases,' and the system evaluation in §4.2 uses a custom 10-cohort suite. No statement in the paper explains how QC-62 relates to the public NeuroBench suite, and no NeuroBench leaderboard or external benchmark results are provided. As written, the headline claim is not verifiable from the manuscript; either report the exact relationship to NeuroBench or change the abstract and conclusion to say 'a held-out internal benchmark (QC-62)'.
  3. [Abstract, §4.2, §5.2] The token-efficiency headline is inconsistent across the paper. The Abstract says 'aggregate token count of about 10.6%' compared to NeuroClaw, but Tables 4 and 6 give total tokens 68,775 vs 2,001,492, i.e., 3.4% of NeuroClaw's aggregate tokens; 10.6% is the per-NIfTI ratio (1,127.5/10,646.2). Section 5.2 itself concedes that per-NIfTI normalization favors NeuroClaw because NeuroClaw emits 188 volumes while CyberNeuro triages to 61. The manuscript should report total-token ratios as primary and per-NIfTI ratios as secondary, and correct the abstract.
minor comments (4)
  1. [Tables 4–6] Each total row contains an extra unlabeled numeric value (33.72, 32.58, 13.65) after 'Time (s)'; presumably seconds per NIfTI, but it is not defined and does not match the table header. Please add a column label or remove the value.
  2. [Figure 4b] The figure states '9.3×' but the table values give 10,646.2/1,127.5 ≈ 9.4. The '89.4% reduction' wording in §4.2 is consistent with 10.6% of baseline, so this is a local arithmetic inconsistency.
  3. [Table 3 / Fig. 2d] Per-category counts (e.g., 11 quality-judgment cases, 6 command-construction cases, 4 escalation cases) are reported in Fig. 2d but not in Table 3. Include the n per category in the table so the abstention-recall and category-level percentages are interpretable.
  4. [§3.7.3] The 'four diagnostic cases' that motivated Round 2 are not described. At minimum, specify whether they come from training data, a development set, or from the QC-62 test set itself; this is directly relevant to the contamination concern in my first major comment.

Circularity Check

0 steps flagged · score 1.0 of 10

No demonstrated circularity: central claims rest on external benchmarks and a held-out evaluation with explicit decontamination; minor self-citations are not load-bearing.

full rationale

The central derivation chain is not circular on the evidence in the text. WandaMind is trained on a described corpus (315 teacher trajectories, 306 rule-based expansions, 101 targeted cases), filtered through schema validation and token-level decontamination against QC-62 (Jaccard >= 0.55), and then evaluated on the held-out QC-62 set. The paper explicitly attempts to prevent leakage rather than constructing the evaluation from the training set: 'To reduce evaluation leakage, we then compare the token set of each candidate with every QC-62 task and remove candidates with Jaccard similarity >= 0.55.' The claimed 40.3% to 69.4% gain is therefore a standard supervised-evaluation result, not a prediction that is forced by fitting QC-62. General-capability checks use external benchmarks (MMLU-Pro, MedQA, PubMedQA, IFEval), and the system-level DICOM-to-BIDS comparison is against the external NeuroClaw system on the public NeuroBench suite; the token-efficiency numbers are arithmetic from the reported tables, with the per-NIfTI normalization and its baseline-favoring effect explicitly disclosed in Section 5.2. The only self-citations, CyberNeuro Assistant (Wu et al., 2026) and the WandaCore agentic harness, appear as related-work or implementation references and are not load-bearing for the paper's main claims. A residual risk exists that semantic near-duplicates below the Jaccard threshold could contaminate the held-out set, but the paper contains no statement that the Round 2/3 hardening counterexamples were derived from QC-62, so treating this as circularity would be speculation rather than a demonstrated reduction. Accordingly, no circular step can be exhibited from the text.

Assumptions & free parameters 3 free parameters · 6 assumptions · 0 invented entities

The central empirical claims rest on the trained WandaMind model and the evaluation methodology; the model itself is a product of supervised learning with hand-chosen hyperparameters, and the benchmark integrity relies on decontamination assumptions.

free parameters (3)
  • LoRA rank and scale (r=32, α=64, α/r=2.0) = r=32, α=64, dropout=0.05
    Hyperparameters chosen by hand for QLoRA fine-tuning; not tuned on a validation set but affect model behavior and thus all evaluation numbers.
  • Jaccard decontamination threshold = 0.55
    Chosen threshold to remove training examples overlapping with QC-62; the threshold determines how much leakage protection is applied.
  • Quantization Q4_K_M = Q4_K_M (62.9% QC-62)
    Deployment quantization chosen to fit memory; the deployed checkpoint scores 6.5 points lower than full precision, so the choice affects reported accuracy.
assumptions (6)
  • domain assumption The QC-62 labels (quality judgment, command construction, escalation) are correct and the 62 cases are representative of real neuroimaging tool-use.
    The central evaluation claims rest on the validity of this self-created benchmark; errors or biases in labeling would directly inflate the reported accuracy.
  • domain assumption The Jaccard decontamination threshold (≥0.55) is sufficient to prevent leakage between training and QC-62.
    The paper assumes token-level similarity removal prevents contamination, but semantic near-duplicates may still overlap; this is load-bearing.
  • domain assumption The base model Qwen2.5-7B-Instruct and its fine-tuned variant retain adequate general capabilities for the workflows tested.
    The system's end-to-end viability assumes the model can interpret planner/validator/consumer roles; IFEval regressions suggest some capability drop.
  • domain assumption The 10 public cohorts (ADNI, AIBL, etc.) are representative of typical DICOM-to-BIDS workloads.
    The token and intervention numbers are measured on these cohorts; other cohorts might show different triage or intervention rates.
  • domain assumption The network-denial measurement methodology (process-level deny-network, netstat deltas) accurately captures outbound traffic for tested workflows.
    The paper itself notes host-wide netstat deltas are not process-attributed, so the zero-egress claim is only supported by the deny-network conditions.
  • domain assumption The threat model assumes the Windows host and bundled WSL are trusted.
    The privacy boundary does not protect against a compromised host or malicious administrator, as stated in §3.2.

how reviews work

0 comments
Cite this review

Pith. "Pith review of CyberNeuro: A Privacy-Preserving Agentic Workbench for Cohort-Scale Neuroimage and Clinical Data Analysis." pith.science (2026). https://pith.science/paper/FTCGOWUD

@misc{pith2026260728841,
  author       = {Pith},
  title        = {Pith review of: CyberNeuro: A Privacy-Preserving Agentic Workbench for Cohort-Scale Neuroimage and Clinical Data Analysis},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/FTCGOWUD}},
  note         = {Machine review of arXiv:2607.28841}
}
read the original abstract

Despite tremendous success in neuroimaging methodology, making large-scale, high-dimensional datasets ready for AI/ML applications remains a critical operational bottleneck. Conventional workflows require extensive manual effort across metadata curation, pipeline execution, post-processing quality control, and data management, a burden that disproportionately excludes laboratories with limited manpower and computational infrastructure. To address this real-world barrier, there is an urgent need for scalable, cost-effective computational platforms that democratize advanced neuroimaging analytics and accelerate discoveries in mental health and clinical translation. Capitalizing on multi-agent LLM breakthroughs, we introduce CyberNeuro, an agentic workbench with a tailored local LLM-model ('WandaMind') for automated neuroimaging and health-data analysis. Driven by four dedicated agents (Planner, Validator, Dispatcher, and Reporter) communicating via a secure MCP bridge and a pinned execution layer, CyberNeuro enables researchers to execute complex workflows using natural language while maintaining clinical-grade data privacy. On the public NeuroBench suite, CyberNeuro increases held-out domain accuracy from 40% to 69% over the baseline model. Beyond automated metrics, the platform integrates a human-in-the-loop verification panel to ensure rigorous biomedical quality control. Across the same end-to-end 10-batch cohort workflow suite, the local WandaMind configuration completed all tasks with an estimated aggregate token count of about 10.6% using WandaMind and 61.7% using cloud providers of token usage, compared to Neuroclaw, respectively. The platform and its production-ready modules are available at https://wanda-cyberbench.com.

Figures

Figures reproduced from arXiv: 2607.28841 by the authors.

Figure 1
Figure 1. System architecture of CyberNeuro. The platform operates within a strict local-first privacy boundary on standalone institutional hardware, ensuring no outbound network traffic. The architecture consists of three main tiers. Interaction Layer: Integrates natural language chat requests, a Skill Hub for managing neuroimaging pipelines (e.g., FreeSurfer, fMRIPrep, CIVET), and interactive co-visualization panels. Orches… view at source ↗
Figure 2
Figure 2. The fine-tuning pipeline and evaluation workflow. Overview of the iterative LoRA adapter training for tool-use optimization. a: Training examples are assembled from teacher￾generated tool traces, targeted cases, and rule-based expansions; grounded using neuroimaging quality-control metrics; validated against tool schemas; and decontaminated against QC-62. b: Qwen2.5-7B-Instruct is adapted using response-only QLoRA w… view at source ↗
Figure 3
Figure 3. User interface of the CyberNeuro workbench. The platform integrates three synchronized panels: the left pane manages natural-language orchestration, session history, and execution environments; the middle pane acts as a unified hub for workspace inputs, processing pipelines, and visualization skills; and the right pane provides a dynamic multi-modal quality control (QC) environment, presenting compatible outputs (e.… view at source ↗
Figures from the paper (1 more)
Figure 4
Figure 4. Figure 4: DICOM conversion and BIDS organization workflow. Panel a illustrate the integrated, single-step pipeline of CyberNeuro compared to the multi-stage pipeline required by NeuroClaw. CyberNeuro automatically inspects and triages local cohort archives, converts retained acq…

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

104 extracted references · 1 linked inside Pith

  1. [1]

    and Katz, Daniel S

    Barker, Michelle and Chue Hong, Neil P. and Katz, Daniel S. and Lamprecht, Anna-Lena and Martinez-Ortiz, Carlos and Psomopoulos, Fotis and Harrow, Jennifer and Castro, Leyla Jael and Gruenpeter, Morane and Martinez, Paula Andrea and Honeyman, Tom , title =. Scientific Data , volume =. 2022 , doi =

    2022

  2. [2]

    Wanda Merged Core: An Open Agentic Harness for Local Tool-Using Assistants , year =

  3. [3]

    Model Context Protocol Specification , year =

  4. [4]

    and Behrens, Timothy E

    Jenkinson, Mark and Beckmann, Christian F. and Behrens, Timothy E. J. and Woolrich, Mark W. and Smith, Stephen M. , title =. NeuroImage , volume =. 2012 , doi =

    2012

  5. [5]

    and Lepage, Claude and Lerch, Jason and Fombonne, Eric and Evans, Alan C

    Ad-Dab'bagh, Yasser and Einarson, Dawn and Lyttelton, Oliver and Muehlboeck, Jens-Sebastian and Mok, Kelvin and Ivanov, Ognen and Vincent, Robert D. and Lepage, Claude and Lerch, Jason and Fombonne, Eric and Evans, Alan C. , title =. Proceedings of the 12th Annual Meeting of the Organization for Human Brain Mapping , year =

  6. [6]

    Statistical Parametric Mapping: The Analysis of Functional Brain Images , publisher =

  7. [7]

    and Leahy, Richard M

    Shattuck, David W. and Leahy, Richard M. , title =. Medical Image Analysis , volume =. 2002 , doi =

    2002

  8. [8]

    , title =

    Cox, Robert W. , title =. Computers and Biomedical Research , volume =. 1996 , doi =

    1996

Show all 104 references
  1. [9]

    Craddock, Cameron and Sikka, Sharad and Cheung, Brian and Khanuja, Ranjeet and Ghosh, Satrajit S. and Yan, Chaogan and Li, Qingyang and Lurie, Daniel and Vogelstein, Joshua and Burns, Randal and Colcombe, Stanley and Mennes, Maarten and Kelly, Clare and Di Martino, Adriana and...

  2. [10]

    Brain Connectivity , volume =

    Whitfield-Gabrieli, Susan and Nieto-Castanon, Alfonso , title =. Brain Connectivity , volume =. 2012 , doi =

    2012

  3. [11]

    and Adali, Tulay and Pearlson, Godfrey D

    Calhoun, Vince D. and Adali, Tulay and Pearlson, Godfrey D. and Pekar, James J. , title =. Human Brain Mapping , volume =. 2001 , doi =

    2001

  4. [12]

    Frontiers in Neuroinformatics , volume =

    Garyfallidis, Eleftherios and Brett, Matthew and Amirbekian, Bagrat and Rokem, Ariel and van der Walt, Stefan and Descoteaux, Maxime and Nimmo-Smith, Ian , title =. Frontiers in Neuroinformatics , volume =. 2014 , doi =

    2014

  5. [13]

    and Wang, Yibao and Fern

    Yeh, Fang-Cheng and Verstynen, Timothy D. and Wang, Yibao and Fern. Deterministic Diffusion Fiber Tracking Improved by Quantitative Anisotropy , journal =. 2013 , doi =

    2013

  6. [14]

    , title =

    Wasserthal, Jakob and Neher, Peter and Maier-Hein, Klaus H. , title =. NeuroImage , volume =. 2018 , doi =

    2018

  7. [15]

    Cardoso, M. Jorge and Li, Wenqi and Brown, Richard and Ma, Nic and Kerfoot, Eric and Wang, Yiheng and Murrey, Benjamin and Myronenko, Andriy and Zhao, Can and Yang, Dong and Nath, Vishwesh and He, Yufan and Xu, Ziyue and Hatamizadeh, Ali and Others , title =. arXiv preprint ar...

    arXiv

  8. [16]

    Computer Methods and Programs in Biomedicine , volume =

    P. Computer Methods and Programs in Biomedicine , volume =. 2021 , doi =

    2021

  9. [17]

    and Auer, Tibor and Calhoun, Vince D

    Gorgolewski, Krzysztof J. and Auer, Tibor and Calhoun, Vince D. and Craddock, R. Cameron and Das, Samir and Duff, Eugene A. and others , title =. Scientific Data , volume =. 2016 , doi =

    2016

  10. [18]

    and Blair, Ross W

    Esteban, Oscar and Markiewicz, Christopher J. and Blair, Ross W. and Moodie, Craig A. and Isik, A. Ilkay and Erramuzpe, Asier and others , title =. Nature Methods , volume =. 2019 , doi =

    2019

  11. [19]

    NeuroImage , volume =

    Fischl, Bruce , title =. NeuroImage , volume =. 2012 , doi =

    2012

  12. [20]

    and He, Xiaosong and Yeh, Fang-Cheng and Dhollander, Thijs and Adebimpe, Azeez and others , title =

    Cieslak, Matthew and Cook, Philip A. and He, Xiaosong and Yeh, Fang-Cheng and Dhollander, Thijs and Adebimpe, Azeez and others , title =. Nature Methods , volume =. 2021 , doi =

    2021

  13. [21]

    and Poldrack, Russell A

    Esteban, Oscar and Birman, Daniel and Schaer, Marie and Koyejo, Oluwasanmi O. and Poldrack, Russell A. and Gorgolewski, Krzysztof J. , title =. PLoS ONE , volume =. 2017 , doi =

    2017

  14. [22]

    Harmonization of Multi-Site Diffusion Tensor Imaging Data , journal =

    Fortin, Jean-Philippe and Parker, Drew and Tun. Harmonization of Multi-Site Diffusion Tensor Imaging Data , journal =. 2017 , doi =

    2017

  15. [23]

    NeuroImage , volume =

    Tournier, J-Donald and Smith, Robert and Raffelt, David and Tabbara, Rami and Dhollander, Thijs and Pietsch, Maximilian and others , title =. NeuroImage , volume =. 2019 , doi =

    2019

  16. [24]

    Frontiers in Neuroinformatics , volume =

    Abraham, Alexandre and Pedregosa, Fabian and Eickenberg, Michael and Gervais, Philippe and Mueller, Andreas and Kossaifi, Jean and others , title =. Frontiers in Neuroinformatics , volume =. 2014 , doi =

    2014

  17. [25]

    and Strohmeier, Daniel and Brodbeck, Christian and others , title =

    Gramfort, Alexandre and Luessi, Martin and Larson, Eric and Engemann, Denis A. and Strohmeier, Daniel and Brodbeck, Christian and others , title =. Frontiers in Neuroscience , volume =. 2013 , doi =

    2013

  18. [26]

    Magnetic Resonance Imaging , volume =

    Fedorov, Andriy and Beichel, Reinhard and Kalpathy-Cramer, Jayashree and Finet, Julien and Fillion-Robin, Jean-Christophe and Pujol, Sonia and others , title =. Magnetic Resonance Imaging , volume =. 2012 , doi =

    2012

  19. [27]

    and Piven, Joseph and Hazlett, Heather Cody and Smith, Rachel Gimpel and Ho, Sean and Gee, James C

    Yushkevich, Paul A. and Piven, Joseph and Hazlett, Heather Cody and Smith, Rachel Gimpel and Ho, Sean and Gee, James C. and Gerig, Guido , title =. NeuroImage , volume =. 2006 , doi =

    2006

  20. [28]

    and Tustison, Nicholas J

    Avants, Brian B. and Tustison, Nicholas J. and Song, Gang and Cook, Philip A. and Klein, Arno and Gee, James C. , title =. NeuroImage , volume =. 2011 , doi =

    2011

  21. [29]

    PLoS ONE , volume =

    Xia, Mingrui and Wang, Jinhui and He, Yong , title =. PLoS ONE , volume =. 2013 , doi =

    2013

  22. [30]

    and Montez, David F

    Marek, Scott and Tervo-Clemmens, Brenden and Calabro, Finnegan J. and Montez, David F. and Kay, Benjamin P. and Hatoum, Alexander S. and others , title =. Nature , volume =. 2022 , doi =

    2022

  23. [31]

    2026 , doi =

    Wang, Cheng and He, Zhibin and Peng, Zhihao and Liu, Shengyuan and Hu, Yufan and Yang, Carl and He, Lifang and Sun, Lichao and Li, Xiang and Yuan, Yixuan , journal =. 2026 , doi =

    2026

  24. [32]

    and Shen, Yelong and Wallis, Phillip and Allen-Zhu, Zeyuan and Li, Yuanzhi and Wang, Shean and Wang, Lu and Chen, Weizhu , booktitle =

    Hu, Edward J. and Shen, Yelong and Wallis, Phillip and Allen-Zhu, Zeyuan and Li, Yuanzhi and Wang, Shean and Wang, Lu and Chen, Weizhu , booktitle =. 2022 , url =

    2022

  25. [33]

    2023 , url =

    Dettmers, Tim and Pagnoni, Artidoro and Holtzman, Ari and Zettlemoyer, Luke , booktitle =. 2023 , url =

    2023

  26. [34]

    2026 , howpublished =

    2026

  27. [35]

    2026 , howpublished =

    Wu, Guorong and Wei, Ziquan and Cao, Shuheng and Hu, Xiyun and Dan, Tingting , title =. 2026 , howpublished =

    2026

  28. [36]

    2018 , howpublished =

    2018

  29. [37]

    Best Practices for Sharing Research Software , howpublished =

  30. [38]

    Advances in Neural Information Processing Systems , volume=

    Brainmoe: Cognition joint embedding via mixture-of-expert towards robust brain foundation model , author=. Advances in Neural Information Processing Systems , volume=

  31. [39]

    The Forty-Third International Conference on Machine Learning (ICML) , year=

    Marrying Generative Model of Healthcare Events with Digital Twin of Social Determinants of Health for Disease Reasoning , author=. The Forty-Third International Conference on Machine Learning (ICML) , year=

  32. [40]

    Advances in Neural Information Processing Systems , volume=

    Let brain rhythm shape machine intelligence for connecting dots on graphs , author=. Advances in Neural Information Processing Systems , volume=

  33. [41]

    Medical Image Analysis , pages=

    Geo-Mamba: Geometry-informed state-space learning of functional brain organization , author=. Medical Image Analysis , pages=. 2026 , publisher=

    2026

  34. [42]

    2026 IEEE 23rd International Symposium on Biomedical Imaging (ISBI) , pages=

    A Physics-Informed and Mechanistically-Grounded Deep Learning Framework for Revealing Altered Structurefunction Coupling in Neurodegenerative Diseases , author=. 2026 IEEE 23rd International Symposium on Biomedical Imaging (ISBI) , pages=. 2026 , organization=

    2026

  35. [43]

    Proceedings of the AAAI Conference on Artificial Intelligence , volume=

    SyncBrain: Exploring Brain Functional Dynamics Through Neural Oscillatory Synchronization , author=. Proceedings of the AAAI Conference on Artificial Intelligence , volume=

  36. [44]

    Medical Image Analysis , pages=

    NeuroDetour: A Neural Pathway Transformer for Uncovering Structural-Functional Coupling Mechanisms in Human Connectome , author=. Medical Image Analysis , pages=. 2026 , publisher=

    2026

  37. [45]

    Journal of the Royal Society Interface , year=

    Optimal Control For Anti-Abeta Treatment in Alzheimer's Disease using a Reaction-Diffusion Model , author=. Journal of the Royal Society Interface , year=

  38. [46]

    The Fourteenth International Conference on Learning Representations , year=

    MnemoDyn: Learning Resting State Dynamics from 40 K FMRI sequences , author=. The Fourteenth International Conference on Learning Representations , year=

  39. [47]

    Conditional diffusion with ordinal regression: Longitudinal data generation for neurodegenerative disease studies , author=

  40. [48]

    International Conference on Medical Image Computing and Computer-Assisted Intervention , pages=

    Understanding brain dynamics through neural koopman operator with structure-function coupling , author=. International Conference on Medical Image Computing and Computer-Assisted Intervention , pages=. 2024 , organization=

    2024

  41. [49]

    dcm2bids — DICOM to BIDS Conversion Skill , year =

  42. [50]

    IEEE Transactions on Biomedical Engineering , volume=

    Scanning the Horizon of Replicability in Neuroscience: A Recipe of Developing Replicable Deep Models for Functional Neuroimages , author=. IEEE Transactions on Biomedical Engineering , volume=. 2025 , publisher=

    2025

  43. [51]

    Proceedings of the AAAI Conference on Artificial Intelligence , volume=

    Large connectome model: An fMRI foundation model of brain connectomes empowered by brain-environment interaction in multitask learning landscape , author=. Proceedings of the AAAI Conference on Artificial Intelligence , volume=

  44. [52]

    International Conference on Information Processing in Medical Imaging , pages=

    A Multi-layer Neural Transport Model for Characterizing Pathology Propagation in Neurodegenerative Diseases , author=. International Conference on Information Processing in Medical Imaging , pages=. 2025 , organization=

    2025

  45. [53]

    International Conference on Medical Image Computing and Computer-Assisted Intervention , pages=

    Brain-Environment Cross-Attention (BECA) Meta-matching: A New Perspective of Brain Connectome Zero-Shot Learning , author=. International Conference on Medical Image Computing and Computer-Assisted Intervention , pages=. 2025 , organization=

    2025

  46. [54]

    International Conference on Medical Image Computing and Computer-Assisted Intervention , pages=

    A Wasserstein Recipe for Replicable Machine Learning on Functional Neuroimages , author=. International Conference on Medical Image Computing and Computer-Assisted Intervention , pages=. 2024 , organization=

    2024

  47. [55]

    2025 IEEE 22nd International Symposium on Biomedical Imaging (ISBI) , pages=

    Understanding the Spreading Mechanism of Tau Propagation in Alzheimers Disease Through a Multi-Layer Transport , author=. 2025 IEEE 22nd International Symposium on Biomedical Imaging (ISBI) , pages=. 2025 , organization=

    2025

  48. [56]

    IEEE transactions on medical imaging , year=

    Understanding Brain Functional Dynamics Through Neural Koopman Operator With Control Mechanism , author=. IEEE transactions on medical imaging , year=

  49. [57]

    International Conference on Medical Image Computing and Computer-Assisted Intervention , pages=

    A General Stitching Solution for Whole-Brain 3D Nuclei Instance Segmentation from Microscopy Images , author=. International Conference on Medical Image Computing and Computer-Assisted Intervention , pages=. 2023 , organization=

    2023

  50. [58]

    International Conference on Information Processing in Medical Imaging , pages=

    Holobrain: A harmonic holography for self-organized brain function , author=. International Conference on Information Processing in Medical Imaging , pages=. 2023 , organization=

    2023

  51. [59]

    Journal of Alzheimer's Disease Reports , volume=

    Uncovering Diverse Mechanistic Spreading Pathways in Disease Progression of Alzheimer’s Disease , author=. Journal of Alzheimer's Disease Reports , volume=. 2023 , publisher=

    2023

  52. [60]

    2023 IEEE 20th International Symposium on Biomedical Imaging (ISBI) , pages=

    Osr-net: Ordinary differential equation-based brain state recognition neural network , author=. 2023 IEEE 20th International Symposium on Biomedical Imaging (ISBI) , pages=. 2023 , organization=

    2023

  53. [61]

    and Jette, Morris A

    Yoo, Andy B. and Jette, Morris A. and Grondona, Mark , title =. Job Scheduling Strategies for Parallel Processing (JSSPP) , series =. 2003 , doi =

    2003

  54. [62]

    , title =

    Hudak, David and Johnson, Doug and Chalker, Alan and Nicklas, Jeremy and Franz, Eric and Dockendorf, Trey and McMichael, Brian L. , title =. Journal of Open Source Software , volume =. 2018 , doi =

    2018

  55. [63]

    and Sochat, Vanessa and Bauer, Michael W

    Kurtzer, Gregory M. and Sochat, Vanessa and Bauer, Michael W. , title =. PLOS ONE , volume =. 2017 , doi =

    2017

  56. [64]

    Nature , volume =

    Autonomous Chemical Research with Large Language Models , author =. Nature , volume =. 2023 , doi =

    2023

  57. [65]

    and Cox, Sam and Schilter, Oliver and Baldassari, Carlo and White, Andrew D

    Bran, Andres M. and Cox, Sam and Schilter, Oliver and Baldassari, Carlo and White, Andrew D. and Schwaller, Philippe , journal =. 2024 , doi =

    2024

  58. [66]

    Brooke, John , booktitle =

  59. [67]

    and Lorenz, Romy and Goncalves, Mathias and MacNicol, Eilidh E

    Ciric, Rastko and Thompson, William H. and Lorenz, Romy and Goncalves, Mathias and MacNicol, Eilidh E. and Markiewicz, Christopher J. and Halchenko, Yaroslav O. and Ghosh, Satrajit S. and Gorgolewski, Krzysztof J. and Esteban, Oscar , journal =. 2022 , doi =

    2022

  60. [68]

    Educational and Psychological Measurement , volume =

    A Coefficient of Agreement for Nominal Scales , author =. Educational and Psychological Measurement , volume =. 1960 , doi =

    1960

  61. [69]

    Dettmers, Tim and Pagnoni, Artidoro and Holtzman, Ari and Zettlemoyer, Luke , booktitle =

  62. [70]

    Medical Image Analysis , volume =

    D. Medical Image Analysis , volume =. 2024 , doi =

    2024

  63. [71]

    Nature Biotechnology , volume =

    Nextflow Enables Reproducible Computational Workflows , author =. Nature Biotechnology , volume =. 2017 , doi =

    2017

  64. [72]

    The Annals of Statistics , volume =

    Bootstrap Methods: Another Look at the Jackknife , author =. The Annals of Statistics , volume =. 1979 , doi =

    1979

  65. [73]

    and Poldrack, Russell A

    Esteban, Oscar and Birman, Daniel and Schaer, Marie and Koyejo, Oluwasanmi O. and Poldrack, Russell A. and Gorgolewski, Krzysztof J. , journal =. 2017 , doi =

    2017

  66. [74]

    2024 , doi =

    Fernandez-Lozano, Sofia and Dadar, Mahsa and Morrison, Cassandra and others , journal =. 2024 , doi =

    2024

  67. [75]

    2023 , url =

    Frantar, Elias and Ashkboos, Saleh and Hoefler, Torsten and Alistarh, Dan , booktitle =. 2023 , url =

    2023

  68. [76]

    and Madison, Cindee and Clark, Dav and Halchenko, Yaroslav O

    Gorgolewski, Krzysztof and Burns, Christopher D. and Madison, Cindee and Clark, Dav and Halchenko, Yaroslav O. and Waskom, Michael L. and Ghosh, Satrajit S. , journal =. 2011 , doi =

    2011

  69. [77]

    and Alfaro-Almagro, Fidel and Auer, Tibor and others , journal =

    Gorgolewski, Krzysztof J. and Alfaro-Almagro, Fidel and Auer, Tibor and others , journal =. 2017 , doi =

    2017

  70. [78]

    Not What You've Signed Up For: Compromising Real-World

    Greshake, Kai and Abdelnabi, Sahar and Mishra, Shailesh and Endres, Christoph and Holz, Thorsten and Fritz, Mario , booktitle =. Not What You've Signed Up For: Compromising Real-World. 2023 , doi =

    2023

  71. [79]

    and Staveland, Lowell E

    Hart, Sandra G. and Staveland, Lowell E. , booktitle =. Development of. 1988 , doi =

    1988

  72. [80]

    Nature Methods , volume =

    brainlife.io: A Decentralized and Open-Source Cloud Platform to Support Neuroscience Research , author =. Nature Methods , volume =. 2024 , doi =

    2024

  73. [81]

    Nature Machine Intelligence , volume =

    Secure, Privacy-Preserving and Federated Machine Learning in Medical Imaging , author =. Nature Machine Intelligence , volume =. 2020 , doi =

    2020

  74. [82]

    NeuroImage , volume =

    Mindcontrol: A Web Application for Brain Segmentation Quality Control , author =. NeuroImage , volume =. 2018 , doi =

    2018

  75. [83]

    and van de Kamp, Femke and van der Meulen, Mara and Peters, Sabine and Wierenga, Lara M

    Klapwijk, Eduard T. and van de Kamp, Femke and van der Meulen, Mara and Peters, Sabine and Wierenga, Lara M. , journal =. 2019 , doi =

    2019

  76. [84]

    Bioinformatics , volume =

    Snakemake---A Scalable Bioinformatics Workflow Engine , author =. Bioinformatics , volume =. 2012 , doi =

    2012

  77. [85]

    PLOS ONE , volume =

    Singularity: Scientific Containers for Mobility of Compute , author =. PLOS ONE , volume =. 2017 , doi =

    2017

  78. [86]

    and Ventola, Pamela and Duncan, James S

    Li, Xiaoxiao and Gu, Yufeng and Dvornek, Nicha and Staib, Lawrence H. and Ventola, Pamela and Duncan, James S. , journal =. Multi-Site. 2020 , doi =

    2020

  79. [87]

    Lin, Ji and Tang, Jiaming and Tang, Haotian and Yang, Shang and Chen, Wei-Ming and Wang, Wei-Chen and Xiao, Guangxuan and Dang, Xingyu and Gan, Chuang and Han, Song , booktitle =

  80. [88]

    2024 , url =

    Liu, Xiao and Yu, Hao and Zhang, Hanchen and Xu, Yifan and Lei, Xuanyu and Lai, Hanyu and Gu, Yu and Ding, Hangliang and Men, Kaiwen and Yang, Kejuan and Zhang, Shudan and Deng, Xiang and Zeng, Aohan and Du, Zhengxiao and Zhang, Chenhui and Shen, Sheng and Zhang, Tianjun and S...

    2024

  81. [89]

    Nature Methods , volume =

    Neurodesk: An Accessible, Flexible and Portable Data Analysis Environment for Reproducible Neuroimaging , author =. Nature Methods , volume =. 2024 , doi =

    2024

  82. [90]

    2023 , doi =

    O'Donoghue, Odhran and Shtedritski, Aleksandar and Ging, Simon and others , booktitle =. 2023 , doi =

    2023

  83. [91]

    and Gorgolewski, Krzysztof J

    Poldrack, Russell A. and Gorgolewski, Krzysztof J. , journal =. The Past, Present, and Future of the Brain Imaging Data Structure (. 2019 , doi =

    2019

  84. [92]

    ACM Computing Surveys , volume =

    Tool Learning with Foundation Models , author =. ACM Computing Surveys , volume =. 2024 , doi =

    2024

  85. [93]

    2025 , doi =

    Qwen Team and others , journal =. 2025 , doi =

    2025

  86. [94]

    2023 , doi =

    Raamana, Pradeep Reddy , journal =. 2023 , doi =

    2023

  87. [95]

    npj Digital Medicine , volume =

    The Future of Digital Health with Federated Learning , author =. npj Digital Medicine , volume =. 2020 , doi =

    2020

  88. [96]

    Proceedings of the IEEE , volume =

    The Protection of Information in Computer Systems , author =. Proceedings of the IEEE , volume =. 1975 , doi =

    1975

  89. [97]

    PLOS Computational Biology , volume =

    Ten Simple Rules for Reproducible Computational Research , author =. PLOS Computational Biology , volume =. 2013 , doi =

    2013

  90. [98]

    Advances in Neural Information Processing Systems , volume =

    Toolformer: Language Models Can Teach Themselves to Use Tools , author =. Advances in Neural Information Processing Systems , volume =

  91. [99]

    Shen, Yiqing and Song, Kaitao and Tan, Xu and Li, Dongsheng and Lu, Weiming and Zhuang, Yueting , booktitle =

  92. [100]

    Advances in Neural Information Processing Systems , volume =

    Reflexion: Language Agents with Verbal Reinforcement Learning , author =. Advances in Neural Information Processing Systems , volume =

  93. [101]

    PeerJ Computer Science , volume =

    Software Citation Principles , author =. PeerJ Computer Science , volume =. 2016 , doi =

    2016

  94. [102]

    Data Science , volume =

    Soiland-Reyes, Stian and Sefton, Peter and Crosas, Merc. Data Science , volume =. 2022 , doi =

    2022

  95. [103]

    npj Digital Medicine , volume =

    Privacy-Preserving Large Language Models for Structured Medical Information Retrieval , author =. npj Digital Medicine , volume =. 2024 , doi =

    2024

  96. [104]

    2023 , url =

    Yao, Shunyu and Zhao, Jeffrey and Yu, Dian and Du, Nan and Shafran, Izhak and Narasimhan, Karthik and Cao, Yuan , booktitle =. 2023 , url =

    2023

Pith tools

Reviewed August 3, 2026 · model on record in the stance chip above.