Pith. sign in

REVIEW 4 major objections 6 minor 61 references

Empowering Functional Neuroimaging: A Pre-trained Generative Framework for Unified Representation of Neural Signals

T0 review · 4 major / 6 minor · reviewed 2026-08-07 · deepseek-v4-flash

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

Pith's one-line read A unified generative model turns cheap EEG into fMRI-like BOLD signals

desk verdict A credible EEG-to-fMRI/fNIRS generation study with one unearned claim: the fairness gains are not benchmarked against trivial oversampling. read the letter →

arxiv 2506.02433 v1 pith:GXQN7P3I submitted 2025-06-03 cs.CV

Weiheng Yao , Xuhang Chen , Shuqiang Wang This is my paper
classification cs.CV
keywords functionalneuroimaginggenerativeAIunifiedrepresentationEEG-to-fMRIsynthesisdiffusionmodelbrain-computerinterfacefairnesscross-modalgeneration
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

Functional neuroimaging that reveals brain activity in detail—fMRI and fNIRS—is expensive, immobile, and hard to acquire, while EEG is cheap and ubiquitous. This paper argues that a pre-trained generative framework can learn a single representation space shared by EEG, fMRI, and fNIRS, and then reconstruct the expensive modality from the cheap one: from EEG alone it generates blood-oxygen-level-dependent (BOLD) time courses that correlate with real BOLD (Pearson correlation 0.43–0.50 versus a noise baseline near 0.03–0.06) and match its spatial structure at a structural similarity score (SSIM) of about 0.81. The same generated data, used as augmentation for an artificially imbalanced dataset, raises the minority-class F1 score from 0.539 ± 0.397 to 0.881 ± 0.040 in an fMRI visual-decoding task and from 0.037 ± 0.052 to 0.743 ± 0.016 in an fNIRS motor-imagery task. If these results hold, advanced functional neuroimaging becomes dramatically cheaper—the authors estimate a 90% reduction in operational cost—and brain-computer interface (BCI) decoding models trained on imbalanced brain data become fairer.

What carries the argument

The unifying mechanism is the 'unified representation space.' It is produced by three coupled components: pre-trained feature extractors for each modality; a hyperdimensional integration step that aligns the two modalities in space and time (electrode signals are re-weighted onto cortical sampling points by inverse squared distance, and time series are aligned by a Gaussian kernel matrix centered at the approximately 6-second hemodynamic delay); and a Diffusion Transformer that learns the joint distribution of the aligned representations. A modal-unpatcher decoder then reconstructs the target modality from a sample of the unified representation. This machinery is what transfers the temporal richness of EEG into the spatial detail of BOLD or fNIRS while keeping the physiological relationship between electrical and hemodynamic signals intact.

What would settle it

The deciding experiment is a cross-site transfer test: train the framework on simultaneous EEG-fMRI data from one scanner or site, generate BOLD for EEG from a second site or scanner with different acquisition parameters, and compare against real BOLD; if the cross-subject Pearson correlation falls to the noise baseline (about 0.03–0.06) or the SSIM drops far below 0.81, the claimed cross-subject and cross-task generalization fails. A complementary check is to apply the augmentation procedure to a dataset with genuine demographic imbalance and see whether the minority-class F1 gain persists.

Watch

Extended reading notes

Core claim

On its own terms, the paper establishes that a single pre-trained generative pipeline—pre-trained feature extractors, a hyperdimensional spatial and temporal alignment step, a Diffusion Transformer generative module, and a modal-unpatcher decoder—can map simultaneously recorded EEG and either fMRI or fNIRS into one unified representation, and can then decode that representation back into the target modality. Trained on paired EEG–fMRI data, the model generates BOLD from EEG alone; the generated BOLD reproduce regional activations, functional-connectivity structure, and the known approximately 6-second hemodynamic delay. Trained on EEG–fNIRS data, it generates fNIRS oxy- and deoxy-hemoglobin signals from EEG. The authors further claim that using these generated samples to rebalance underrepresented classes reduces the performance gap between minority and majority groups, and that the unified representations support downstream decoding—visual stimulus, motor imagery, body-mass-index and age regression—at levels close to or beyond what real BOLD provides.

Load-bearing premise

The load-bearing premise is that a mapping learned from small paired datasets of roughly 17–29 subjects captures enough population variability that the same EEG-to-BOLD and EEG-to-fNIRS mapping works across new subjects, tasks, and clinical groups; the paper's own limitations acknowledge that most available datasets have fewer than 30 subjects and that overfitting to dominant patterns occurs.

Editorial extensions

If this is right

  • fMRI and fNIRS signals could be synthesized from EEG alone in settings where only EEG recording is feasible, removing the primary cost and mobility barriers to multimodal functional neuroimaging.
  • Balancing an imbalanced BCI training set with generated samples should raise the performance of underrepresented classes, reduce prediction variance, and shrink the accuracy gap with well-represented classes.
  • Downstream decoding—visual stimulus identity, motor imagery state, and continuous physiological prediction—should reach levels close to those obtained with real BOLD, and in some cases (for example BMI prediction) surpass them.
  • The generated signals carry interpretable neurophysiological structure, including the approximately 6-second hemodynamic lag, gamma-band dominance in visual tasks, and functional connectivity patterns overlapping with real BOLD.

Reading between the lines

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

  • A natural reverse test the paper does not run: if the unified representation space is truly shared, the framework should also reconstruct EEG from fMRI or fNIRS; the paper only demonstrates cheap-to-expensive synthesis, so a symmetric experiment would probe whether the space is genuinely modality-invariant.
  • The fairness improvement is demonstrated on simulated task-level imbalance within small datasets, not on demographic imbalance; applying the same augmentation to a dataset with real group underrepresentation (by age, sex, or clinical status) would show whether the fairness claim extends beyond the simulated setting.
  • The temporal correlations reported (0.43–0.50) are modest in absolute terms, which suggests the framework's main added value may be spatial and structural fidelity (SSIM of about 0.81 and connectivity preservation) rather than high-fidelity time-course reconstruction; a comparison against a simple delayed linear regression on the same surfaces would settle this.
  • The paper's own limitation note—that most paired datasets have fewer than 30 subjects—makes cross-site generalization the most likely failure point; a held-out-site or cross-scanner evaluation would be the strongest test of the framework's practical promise.
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. The paper proposes a pre-trained generative framework that maps EEG into fMRI-like BOLD and fNIRS signals using pretrained feature extractors, a hyperdimensional integration module, and a Diffusion Transformer-based unified representation module. The authors evaluate the framework on EEG-fMRI and EEG-fNIRS datasets, reporting moderate temporal correlation (PCC 0.43–0.50 vs. noise 0.03–0.06), spatial similarity (SSIM approximately 0.81), cross-modal and cross-subject generalization, improvements in downstream decoding and clinical decision-support tasks, and fairness gains through data augmentation for underrepresented classes. The main claims are that the framework generates data consistent with real brain activity, provides interpretable brain-mechanism insights, lowers neuroimaging cost, and enhances fairness of BCI decoding models.

Significance. If the central claims hold, the framework would be a practically useful contribution to low-cost neuroimaging and BCI fairness: it provides a concrete pipeline from EEG to hemodynamic modalities, reports quantitative fidelity metrics against noise baselines, and demonstrates a plausible use of synthetic data for class rebalancing. The use of held-out test subjects and multiple datasets is a strength, as is the explicit comparison against a noise baseline for the generation fidelity metrics. However, the current evidence is insufficient to support several headline claims, particularly the fairness contribution, the 90% cost reduction, and the neurobiological interpretability of the SHAP analyses. The novelty relative to the authors' own prior CATD and SCDM methods is also not demonstrated, since no direct comparison is provided.

major comments (4)
  1. [Cross-Modal Data Augmentation Enhances Fairness of BCI Decoding Models (Table 1)] The fairness evaluation lacks any non-generative control. The reported improvements (e.g., Task 5 F1 from 0.539±0.397 to 0.881±0.040; LMI from 0.037±0.052 to 0.743±0.016) are compared only against the imbalanced baseline. Without comparing with simple oversampling of the original 30 minority samples, class reweighting, or SMOTE, the gains could be fully explained by generic class-balance correction, and the specific contribution of the proposed unified representation and generative module is not established. Furthermore, no comparison is made with the authors' own prior EEG-to-fMRI/fNIRS models (refs 30 and 31, CATD and SCDM), so the advantage of the proposed framework over existing unified-representation approaches is unverified. Because fairness enhancement is a central claim of the abstract and the Results, these control experiments are necessary before the claim can be accepted.
  2. [Results, paragraph after Fig. 5] The statement that the framework can 'reduce operational costs by an estimated 90%' is an unsupported quantitative claim; no cost model, calculation, or citation is provided. This figure appears in both the Results and the Discussion, and it substantially overstates what the current experiments can demonstrate. The abstract's use of 'accurately generates' is also stronger than the evidence supports: the PCC values of 0.43–0.50 and SSIM of 0.81 indicate moderate fidelity, and the paper's own Extended Data Fig. 1 documents regional and temporal mismatches. The cost-reduction figure should be removed or replaced with a derived estimate, and the accuracy language should be tempered.
  3. [The Unified Representations Provide Insights into Brain Mechanisms (Fig. 3)] The SHAP analyses are presented as evidence that the framework 'interprets brain mechanisms' and 'captures biologically plausible representations,' but SHAP values on the model's own inputs and outputs measure the sensitivity of the learned mapping, not neurobiological causality. The overlap between SHAP-identified regions of real and generated BOLD and the dominance of the gamma band are consistent with known phenomena, but they do not by themselves establish that the framework provides mechanistic insight. For example, the gamma-band result could arise from the EEG feature extractor's encoding rather than from a true neurovascular coupling learned by the model. The authors should either validate these SHAP-based findings against independent neuroimaging evidence (e.g., known task-evoked networks or prior fMRI studies) or substantially qualify the mechanistic language used in the text and figure captions.
  4. [The Proposed Framework Enables Cross-Modal and Cross-Subject Generalization (Fig. 4(c,d))] The cross-subject generalization claim rests on only three held-out subjects in Fig. 4(c,d), with no confidence intervals or statistical test across subjects, while the training sets contain 17–29 subjects. The paper's own Limitations section admits that 'most available datasets include fewer than 30 subjects' and that the model suffers from 'overfitting to dominant patterns' (Extended Data Fig. 1). Given this, the claim of 'strong generalization across modalities and subjects' is not supported by the scale or statistical treatment of the evidence. The authors should either add more subjects and report per-subject variability with appropriate statistics, or moderate the claim to something like 'preliminary evidence of cross-subject consistency on a small held-out set.'
minor comments (6)
  1. [Fig. 5 caption] The caption contains a typo: 'Simliar' should be 'Similar.'
  2. [Fig. 4 caption] The caption contains a typo: 'revelance' should be 'relevance.'
  3. [Methods, paragraph on organization] The sentence 'The Methods section is organized as follows: First, ... Next, ... Finally, ... This framework provides a comprehensive exposition of the framework, from data preparation to its operationalization and application.' is repetitive and should be rewritten for clarity.
  4. [Eq. (3) and Preprocessing] The physiological delay τ is fixed to approximately 6 seconds both in the preprocessing (BOLD shifted by 6 s) and in Eq. (3). Please clarify whether τ is a fixed hyperparameter in all experiments and whether the reported PCC values are computed after the same shift; otherwise the temporal-consistency result may be partly an artifact of the alignment procedure.
  5. [Methods, Preprocessing and Fig. 2] The 'noise baseline' used for PCC comparisons is never defined in the Methods. Please specify how the noise signals were generated (e.g., Gaussian white noise, phase-randomized surrogate, or shuffled real signals), as the validity of the noise comparison depends on this choice.
  6. [General scope] The abstract and Fig. 1 mention ECoG and fPAI as modalities within the unified representation space, but the experiments cover only EEG, fMRI, and fNIRS. Please clarify explicitly that ECoG and fPAI generation are future directions, not results of this work.

Circularity Check

1 steps flagged · score 2.0 of 10

One temporal-alignment result is circular by construction, but the core held-out generation and decoding experiments are not.

  1. self definitional [Methods ('Preprocessing' and 'Hyperdimensional Integration Strategy', Eq. (3)) vs. Results 'Temporal relationship analysis' (Fig. 2(d))]
    "To account for the physiological delay between electrical brain activity and BOLD signals–typically a 6-second lag [60]–data were segmented and temporally aligned accordingly. Specifically, the onset of blood oxygen signals was shifted to occur 6 seconds after the corresponding electrical activity, enabling the model to accurately capture the temporal relationship between EEG and BOLD responses for precise cross-modal representation."

    The 6-second lag is put into the training data before model training, and the alignment matrix T in Eq. (3) explicitly uses τ approximately 6 seconds. The later Results claim that 'the correlation peaked at -6 seconds' and that this 'aligns with the known hemodynamic delay' is therefore a re-statement of the preprocessing/label shift, not an independent discovery from the generated signals. The model was trained on EEG-BOLD pairs that were already shifted so that BOLD onset follows EEG by 6 seconds, so observing the peak at -6 seconds is largely by construction rather than empirical validation of the learned temporal mapping.

full rationale

Aside from the 6-second temporal-alignment check, the paper's main derivation chain is self-contained. BOLD and fNIRS generation are evaluated on held-out subjects and test segments, with generated signals compared against real signals (PCC, SSIM, noise baselines), so those claims do not reduce to training targets. The fairness augmentation experiment is an empirical downstream evaluation on real test data; whether a simple oversampling baseline would achieve the same F1 gains is an experimental-control question, not a circularity in the paper's equations. Self-citations to the authors' prior SCDM/CATD work are contextual and are not used to justify the framework's validity or uniqueness. The one genuinely circular element is the temporal-delay analysis, where a preprocessing shift of 6 seconds is later reported as a model-discovered hemodynamic delay. This is a minor self-definitional result that does not affect the core generation or decoding claims, so the overall circularity score is low.

Assumptions & free parameters 6 free parameters · 5 assumptions · 0 invented entities

The central method depends on several hand-set temporal and spatial alignment constants, plus design choices in the fairness experiments. No new physical entities are introduced. The main assumptions are about the learnability and physiological validity of the EEG-to-hemodynamic mapping.

free parameters (6)
  • Physiological delay tau = 6 seconds
    Fixed temporal lag used in the time-alignment matrix Eq. (3); if the true lag varies by region or subject, the alignment is biased.
  • Gaussian kernel width sigma = 2 seconds
    Width parameter in Eq. (3) representing the distribution of temporal lag; chosen by hand, not fitted.
  • Regularization epsilon = 1e-6
    Added in the denominator of the inverse-distance weights in Eq. (2) to avoid division by zero.
  • Downsampled mesh vertices = 2562 per hemisphere
    fMRI surface sampling resolution used in preprocessing; affects the spatial detail of generated BOLD.
  • Augmentation sample counts = 120 (fMRI), 90 (fNIRS)
    Numbers of generated samples used to rebalance classes in the fairness experiments; chosen by designers and directly determine the reported fairness improvements.
  • Bandpass filter ranges = EEG 1-100 Hz; fNIRS 0.01-0.1 Hz
    Preprocessing filters; choices affect which signal components are available for learning the cross-modal mapping.
assumptions (5)
  • domain assumption EEG and BOLD/fNIRS are linked by a learnable, subject-independent mapping within the paired training distribution.
    The entire generative framework assumes paired EEG-hemodynamic data can teach a mapping that generalizes across subjects and tasks. Invoked throughout training (Eq. 8).
  • domain assumption The hemodynamic response lags electrical activity by a fixed 6 seconds with Gaussian spread.
    Eq. (3) fixes tau=6s, sigma=2s; the paper's own limitations acknowledge this fails for regional and task-specific variations.
  • domain assumption Spatial correspondence between scalp EEG electrodes and cortical surface vertices can be approximated by inverse-distance-squared weighting.
    Eqs. (1)-(2) in the Hyperdimensional Integration Strategy; no anatomical validation of this weighting is provided.
  • domain assumption SHAP attributions computed on a downstream brain-decoding model reflect underlying neurophysiological mechanisms.
    Figure 3 interprets SHAP values on real and generated BOLD as insights into brain mechanisms.
  • ad hoc to paper Synthetic data generated by the same framework, when added to a classifier's training set, improves fairness on real underrepresented groups without introducing harmful biases.
    Fairness experiments simulate imbalance by downsampling a well-represented task and then augment; no validation is performed on genuinely underrepresented populations.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Empowering Functional Neuroimaging: A Pre-trained Generative Framework for Unified Representation of Neural Signals." pith.science (2026). https://pith.science/paper/GXQN7P3I

@misc{pith2026250602433,
  author       = {Pith},
  title        = {Pith review of: Empowering Functional Neuroimaging: A Pre-trained Generative Framework for Unified Representation of Neural Signals},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/GXQN7P3I}},
  note         = {Machine review of arXiv:2506.02433}
}
read the original abstract

Multimodal functional neuroimaging enables systematic analysis of brain mechanisms and provides discriminative representations for brain-computer interface (BCI) decoding. However, its acquisition is constrained by high costs and feasibility limitations. Moreover, underrepresentation of specific groups undermines fairness of BCI decoding model. To address these challenges, we propose a unified representation framework for multimodal functional neuroimaging via generative artificial intelligence (AI). By mapping multimodal functional neuroimaging into a unified representation space, the proposed framework is capable of generating data for acquisition-constrained modalities and underrepresented groups. Experiments show that the framework can generate data consistent with real brain activity patterns, provide insights into brain mechanisms, and improve performance on downstream tasks. More importantly, it can enhance model fairness by augmenting data for underrepresented groups. Overall, the framework offers a new paradigm for decreasing the cost of acquiring multimodal functional neuroimages and enhancing the fairness of BCI decoding models.

Discussion (0). Sign in to comment.

Reference graph

Works this paper leans on

61 extracted references · 56 canonical work pages

  1. [1]

    Gordon, E. M. et al. A somato-cognitive action network alternates with effector regions in motor cortex. Nature 617, 351–359 (2023)

    work page 2023

  2. [2]

    Jia, G. et al. Modulating emotional states of rats through a rat-like robot with learned interaction patterns. Nature Machine Intelli- gence 6, 1580–1593 (2024)

    work page 2024

  3. [3]

    Endo, M. et al. Data-driven discovery of movement-linked heterogeneity in neurode- generative diseases. Nature Machine Intelli- gence 6, 1034–1045 (2024)

    work page 2024

  4. [4]

    Lin, A. et al. Imaging whole-brain activ- ity to understand behaviour. Nature Reviews Physics 4, 292–305 (2022). 14

    work page 2022

  5. [5]

    Na, S. et al. Massively parallel functional photoacoustic computed tomography of the human brain. Nature Biomedical Engineering 6, 584–592 (2022)

    work page 2022

  6. [6]

    & Kuzum, D

    Ramezani, M., Ren, Y., Cubukcu, E. & Kuzum, D. Innovating beyond electrophys- iology through multimodal neural interfaces. Nature Reviews Electrical Engineering 1–16 (2024)

    work page 2024

  7. [7]

    Bhargava, A. et al. Vascuviz: a multimodal- ity and multiscale imaging and visualization pipeline for vascular systems biology. Nature methods 19, 242–254 (2022)

    work page 2022

  8. [8]

    Logothetis, N. K. What we can do and what we cannot do with fmri. Nature 453, 869–878 (2008)

    work page 2008

Show all 61 references
  1. [9]

    & van der Schaar, M

    van Breugel, B., Liu, T., Oglic, D. & van der Schaar, M. Synthetic data in biomedicine via generative artificial intelligence. Nature Reviews Bioengineering 2, 991–1004 (2024)

    2024

  2. [10]

    Zhang, F. et al. Towards fairness-aware and privacy-preserving enhanced collabora- tive learning for healthcare. Nature Commu- nications 16, 2852 (2025)

    2025

  3. [11]

    Karargyris, A. et al. Federated benchmarking of medical artificial intelligence with med- perf. Nature machine intelligence 5, 799–810 (2023)

    2023

  4. [12]

    Ktena, I. et al. Generative models improve fairness of medical classifiers under distribu- tion shifts. Nature Medicine 30, 1166–1173 (2024)

    2024

  5. [13]

    & Shengye, H

    Wang, S., Yu, W., Chenchen, X. & Shengye, H. Visualization method for evaluating brain addiction traits, apparatus, and medium (2024). US Patent 12,093,833

    2024

  6. [14]

    Wang, S., Yu, W., Chen, Z. et al. Smart diagnosis assistance method to solve results of inaccurate classification of image, and ter- minal based on medical images (2025). US Patent 12,254,684

    2025

  7. [15]

    Qiao, C. et al. Evaluation and development of deep neural networks for image super- resolution in optical microscopy. Nature meth- ods 18, 194–202 (2021)

    2021

  8. [16]

    Zhang, Y. et al. Rapid detection of neurons in widefield calcium imaging datasets after train- ing with synthetic data. Nature Methods 20, 747–754 (2023)

    2023

  9. [17]

    & Zhang, W

    Wang, S., Yanyan, S. & Zhang, W. Enhanced generative adversarial network and target sample recognition method (2024). US Patent 12,154,036

    2024

  10. [18]

    & Wang, S

    Kong, H., Pan, J., Shen, Y. & Wang, S. Adversarial learning based structural brain- network generative model for analyzing mild cognitive impairment. In Chinese Conference on Pattern Recognition and Computer Vision 361–375 (PRCV 2022)

    2022

  11. [19]

    Jing, C. et al. Estimating addiction-related brain connectivity by prior-embedding graph generative adversarial networks. IEEE Trans- actions on Cybernetics 54, 5026–5039 (2024)

    2024

  12. [20]

    & Wang, S

    Gong, C., Chen, X., Mughal, B. & Wang, S. Addictive brain-network identification by spa- tial attention recurrent network with feature selection. Brain Informatics 10, 2 (2023)

    2023

  13. [21]

    Jing, C. et al. Addiction-related brain net- works identification via graph diffusion recon- struction network. Brain Informatics 11, 1 (2024)

    2024

  14. [22]

    Huang, Q. et al. Replay-triggered brain-wide activation in humans. Nature Communica- tions 15, 7185 (2024)

    2024

  15. [23]

    Yu, Y. et al. Sleep fmri with simultaneous electrophysiology at 9.4 t in male mice.Nature communications 14, 1651 (2023)

    2023

  16. [24]

    Moor, M. et al. Foundation models for gen- eralist medical artificial intelligence. Nature 616, 259–265 (2023)

    2023

  17. [25]

    Zhang, K. et al. A generalist vision–language foundation model for diverse biomedical tasks. Nature Medicine 1–13 (2024)

    2024

  18. [26]

    & Shung, D

    Giuffr` e, M. & Shung, D. L. Harnessing the power of synthetic data in healthcare: inno- vation, application, and privacy. NPJ digital medicine 6, 186 (2023)

    2023

  19. [27]

    & Wang, S

    You, S., Shen, Y., Wu, G. & Wang, S. Brain MR images super-resolution with the consis- tent features. In International Conference on Machine Learning and Computing 501–506 (ICMLC ’22)

  20. [28]

    Wang, S. et al. Generative AI enables EEG super-resolution via spatio-temporal adaptive diffusion learning. IEEE Transactions on Consumer Electronics DOI–10 (2025). 15

    2025

  21. [29]

    Li, Y. et al. Generative AI enables the detec- tion of autism using EEG signals. In Chinese Conference on Biometric Recognition 375–384 (CCBR 2023)

    2023

  22. [31]

    Yao, W. et al. CATD: Unified representation learning for EEG-to-fMRI cross-modal gener- ation. IEEE Transactions on Medical Imaging (2025). https://doi.org/10.1109/TMI.2025. 3550206

    2025 doi

  23. [32]

    Wang, J. et al. Self-improving generative foundation model for synthetic medical image generation and clinical applications. Nature Medicine 31, 609–617 (2025)

    2025

  24. [33]

    Mahmood, M. et al. Fully portable and wireless universal brain–machine interfaces enabled by flexible scalp electronics and deep learning algorithm. Nature Machine Intelli- gence 1, 412–422 (2019)

    2019

  25. [34]

    & Halgren, E

    Destrieux, C., Fischl, B., Dale, A. & Halgren, E. Automatic parcellation of human corti- cal gyri and sulci using standard anatomical nomenclature. Neuroimage 53, 1–15 (2010)

    2010

  26. [35]

    Lundberg, S. M. & Lee, S.-I. in A uni- fied approach to interpreting model predictions (eds Guyon, I. et al. ) Advances in Neural Information Processing Systems 30 4765–4774 (Curran Associates, Inc., 2017)

    2017

  27. [36]

    & Kulisevsky, J

    Pagonabarraga, J., Bejr-Kasem, H., Martinez- Horta, S. & Kulisevsky, J. Parkinson disease psychosis: From phenomenology to neurobio- logical mechanisms. Nature Reviews Neurol- ogy 20, 135–150 (2024)

    2024

  28. [37]

    Haghi, B. et al. Enhanced control of a brain– computer interface by tetraplegic participants via neural-network-mediated feature extrac- tion. Nature Biomedical Engineering 1–18 (2024)

    2024

  29. [38]

    Willsey, M. S. et al. A high-performance brain–computer interface for finger decoding and quadcopter game control in an individual with paralysis. Nature Medicine 1–9 (2025)

    2025

  30. [39]

    Womelsdorf, T., Fries, P., Mitra, P. P. & Des- imone, R. Gamma-band synchronization in visual cortex predicts speed of change detec- tion. Nature 439, 733–736 (2006)

    2006

  31. [40]

    Degenhart, A. D. et al. Stabilization of a brain–computer interface via the alignment of low-dimensional spaces of neural activ- ity. Nature biomedical engineering 4, 672–685 (2020)

    2020

  32. [41]

    Chen, X. et al. A neural speech decod- ing framework leveraging deep learning and speech synthesis. Nature Machine Intelligence 1–14 (2024)

    2024

  33. [42]

    Steyaert, S. et al. Multimodal data fusion for cancer biomarker discovery with deep learn- ing. Nature machine intelligence 5, 351–362 (2023)

    2023

  34. [43]

    Liang, J. et al. Deep learning supported dis- covery of biomarkers for clinical prognosis of liver cancer. Nature Machine Intelligence 5, 408–420 (2023)

    2023

  35. [44]

    Liu, X. et al. A generalist medical language model for disease diagnosis assistance. Nature Medicine 1–11 (2025)

    2025

  36. [45]

    Zheng, Q. et al. Large-scale long-tailed dis- ease diagnosis on radiology images. Nature Communications 15, 10147 (2024)

    2024

  37. [46]

    Chen, R. J. et al. Algorithmic fairness in artificial intelligence for medicine and health- care. Nature biomedical engineering 7, 719– 742 (2023)

    2023

  38. [47]

    Bzdok, D., Nichols, T. E. & Smith, S. M. Towards algorithmic analytics for large-scale datasets. Nature Machine Intelligence 1, 296–306 (2019)

    2019

  39. [48]

    Y., Kay, K., Naselaris, T., Tarr, M

    Wang, A. Y., Kay, K., Naselaris, T., Tarr, M. J. & Wehbe, L. Better models of human high-level visual cortex emerge from natu- ral language supervision with a large and diverse dataset. Nature Machine Intelligence 5, 1415–1426 (2023)

    2023

  40. [49]

    Cao, M. et al. Virtual intracranial eeg sig- nals reconstructed from meg with potential for epilepsy surgery. Nature communications 13, 994 (2022)

    2022

  41. [50]

    Pai, S. et al. Foundation model for cancer imaging biomarkers. Nature machine intelli- gence 6, 354–367 (2024). 16

    2024

  42. [51]

    & J¨ ustel, D

    Dehner, C., Zahnd, G., Ntziachristos, V. & J¨ ustel, D. A deep neural network for real- time optoacoustic image reconstruction with adjustable speed of sound. Nature Machine Intelligence 5, 1130–1141 (2023)

    2023

  43. [52]

    Peebles, W. S. & Xie, S. Scalable diffusion models with transformers. IEEE International Conference on Computer Vision (2022)

    2022

  44. [53]

    Telesford, Q. K. et al. An open-access dataset of naturalistic viewing using simultaneous eeg- fmri. Scientific Data 10, 554 (2023)

    2023

  45. [54]

    W., Zhang, G

    Deligianni, F., Carmichael, D. W., Zhang, G. H., Clark, C. A. & Clayden, J. D. Noddi and tensor-based microstructural indices as predictors of functional connectivity. Plos one 11, e0153404 (2016)

    2016

  46. [55]

    Deligianni, F., Centeno, M., Carmichael, D. W. & Clayden, J. D. Relating resting-state fmri and eeg whole-brain connectomes across frequency bands. Frontiers in neuroscience 8, 98767 (2014)

    2014

  47. [56]

    jcavanagh@unm.edu, J. F. C. Eeg: 3-stim auditory oddball and rest in parkinson’s. Available at OpenNeuro (2021). URL https:// doi.org/10.18112/openneuro.ds003490.v1.1.0

    2021 doi

  48. [57]

    Shin, J. et al. Open access dataset for eeg+ nirs single-trial classification. IEEE Trans- actions on Neural Systems and Rehabilitation Engineering 25, 1735–1745 (2016)

    2016

  49. [58]

    J., Rosas, H

    Reuter, M., Schmansky, N. J., Rosas, H. D. & Fischl, B. Within-subject template estima- tion for unbiased longitudinal image analysis. Neuroimage 61, 1402–1418 (2012)

    2012

  50. [59]

    Pang, J. C. et al. Geometric constraints on human brain function. Nature 618, 566–574 (2023)

    2023

  51. [60]

    K., Pauls, J., Augath, M., Trinath, T

    Logothetis, N. K., Pauls, J., Augath, M., Trinath, T. & Oeltermann, A. Neurophysio- logical investigation of the basis of the fmri signal. nature 412, 150–157 (2001)

    2001

  52. [61]

    & liang Lu, B

    Jiang, W., Zhao, L. & liang Lu, B. Large brain model for learning generic represen- tations with tremendous EEG data in BCI (2024)

    2024

  53. [62]

    Sudlow, C. L. M. et al. Uk biobank: An open access resource for identifying the causes of a wide range of complex diseases of middle and old age. PLoS Medicine 12 (2015). 17 Real�BOLD Generated�BOLD Real�BOLD�and�Generated�BOLD�Sequences    ...

    2015

Pith tools

Reviewed August 7, 2026 · model on record in the stance chip above.