pith. machine review for the scientific record. sign in

arxiv: 2605.02500 · v1 · submitted 2026-05-04 · 💻 cs.LG · cs.AI

Recognition: 2 theorem links

Pretraining on Sleep Data Improves non-Sleep Biosignal Tasks

William Lehn-Schi{\o}ler , Magnus Ruud Kj{\ae}r , Phillip Hempel , Magnus Guldberg Pedersen , Rahul Thapa , Bryan He , Nicolai Spicher , Andreas Brink-Kjaer
show 2 more authors
Lars Kai Hansen Emmanuel Mignot
Authors on Pith no claims yet

Pith reviewed 2026-05-08 19:18 UTC · model grok-4.3

classification 💻 cs.LG cs.AI
keywords sleep pretrainingbiosignalsEEGECGcontrastive learningtransfer learningfoundation modelspolysomnography
0
0 comments X

The pith

Sleep biosignal pretraining improves performance on non-sleep EEG and ECG tasks.

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

The paper tests whether pretraining solely on sleep polysomnography data can produce representations that transfer to non-sleep biosignal problems. It applies multimodal contrastive pretraining with a leave-one-out objective on sleep recordings only, then evaluates fine-tuned models on eight downstream EEG and ECG tasks from multiple datasets. Sleep pretraining consistently beats training from scratch and matches or exceeds prior specialized state-of-the-art and foundation models on several tasks. A reader would care because EEG and ECG are widely used in clinical diagnostics, and abundant sleep data could serve as a cheap way to bootstrap better models without needing equally large labeled datasets for each new task.

Core claim

Sleep-only multimodal contrastive pretraining with a leave-one-out objective on polysomnography data yields representations that, after fine-tuning, improve performance relative to training from scratch across eight non-sleep EEG and ECG downstream tasks spanning multiple datasets and, on several tasks, reach or surpass the results of prior specialized state-of-the-art and foundation models.

What carries the argument

Multimodal contrastive pretraining restricted to sleep data and using a leave-one-out objective, which learns shared representations across sleep biosignals that transfer to awake-state EEG and ECG.

Load-bearing premise

The observed performance gains are caused by the sleep pretraining distribution itself rather than differences in model architecture, training duration, or dataset characteristics.

What would settle it

An experiment that trains equivalent models from scratch using the same architecture, total compute, and data volume as the sleep-pretrained versions and finds no consistent improvement or even worse results on the eight downstream tasks would falsify the benefit of sleep pretraining.

Figures

Figures reproduced from arXiv: 2605.02500 by Andreas Brink-Kjaer, Bryan He, Emmanuel Mignot, Lars Kai Hansen, Magnus Guldberg Pedersen, Magnus Ruud Kj{\ae}r, Nicolai Spicher, Phillip Hempel, Rahul Thapa, William Lehn-Schi{\o}ler.

Figure 1
Figure 1. Figure 1: Overview of the study. Left: Sleep PSG is used for self-supervised pretraining of SleepFM1. Right: The pretrained encoder is fine-tuned on eight downstream tasks spanning five EEG and three ECG datasets. Class distributions for each downstream dataset are shown; EEG datasets are evaluated on 60-second windows, while ECG datasets use 10-second segments view at source ↗
Figure 2
Figure 2. Figure 2: SleepFM1 performance without pretraining (w/o PT) and with pretraining (PT) for ECG benchmarks, as well as with pretraining and positional encodings (PT + PE) across EEG benchmarks. Error bars denote 95% confidence intervals: for EEG, intervals are computed across 15 independently trained models; for ECG, intervals are computed across the four arrhythmia detection tasks view at source ↗
Figure 3
Figure 3. Figure 3: SleepFM1 performance with pretraining scaled by the number of input channels on the EEG and ECG datasets. Red points denote when the performance is at or has surpassed 95% of the performance on all channels. particularly notable because LaBraM and LUNA were pre￾trained on the TUH EEG corpus, which includes TUAB recordings — giving them direct domain overlap with the test set that our model does not share. … view at source ↗
read the original abstract

Sleep foundation models have recently demonstrated strong performance on in-domain polysomnography tasks, including sleep staging, apnea detection, and disease risk prediction. In this work, we investigate whether sleep biosignals can serve as an effective pretraining distribution for learning representations that transfer beyond sleep to adjacent domains. Following sleep foundation models, we perform sleep-only multimodal contrastive pretraining (with a leave-one-out objective) and evaluate transfer to non-sleep EEG and ECG, two well-benchmarked biosignal modalities with heterogeneous datasets and clinically meaningful downstream tasks. Across eight downstream tasks spanning multiple EEG and ECG datasets, sleep pretraining consistently improves performance relative to training from scratch. Moreover, on several tasks, we achieve performance competitive with or surpassing prior specialized state-of-the-art and foundation models.

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, simulated authors' rebuttal, and a circularity audit. Tearing a paper down is the easy half of reading it; the pith above is the substance, this is the friction.

Referee Report

3 major / 2 minor

Summary. The manuscript claims that sleep-only multimodal contrastive pretraining (with leave-one-out objective) on polysomnography data yields transferable representations that improve performance on eight non-sleep EEG and ECG downstream tasks relative to training from scratch. On several of these tasks the approach is reported to match or exceed prior specialized state-of-the-art and foundation models.

Significance. If the performance deltas can be isolated to the sleep pretraining distribution after proper controls, the result would indicate that sleep biosignals constitute an effective general-purpose pretraining corpus for biosignal foundation models. This could support more data-efficient learning pipelines in EEG/ECG analysis and reduce the need for large task-specific labeled sets.

major comments (3)
  1. [§4] §4 (Experimental Setup): The description of the from-scratch baselines does not state whether they receive the same total number of optimization steps, epochs, or compute budget as the pretrain-then-fine-tune pipeline. Without this control the observed gains cannot be attributed to the sleep data distribution rather than differences in training duration or hyperparameter effort, which is load-bearing for the central claim.
  2. [§5] §5 (Results): No ablation is presented that trains scratch models for an extended number of steps matching the pretraining plus fine-tuning budget. Such an ablation is required to rule out optimization artifacts as the source of the reported improvements across the eight tasks.
  3. [Table 2] Table 2 or equivalent results table: The manuscript supplies no statistical significance tests, confidence intervals, or error bars on the performance deltas, making it impossible to determine whether the claimed 'consistent improvements' are robust.
minor comments (2)
  1. [Abstract] The abstract would benefit from one or two concrete numerical examples (e.g., accuracy or F1 deltas) to convey the magnitude of the reported gains.
  2. [Methods] The leave-one-out contrastive objective would be clearer if accompanied by an explicit equation or pseudocode in the methods section.

Simulated Author's Rebuttal

3 responses · 0 unresolved

We thank the referee for their constructive comments, which have helped us improve the rigor of our experimental controls and statistical reporting. We address each major comment below and have revised the manuscript accordingly.

read point-by-point responses

  1. Referee: §4 (Experimental Setup): The description of the from-scratch baselines does not state whether they receive the same total number of optimization steps, epochs, or compute budget as the pretrain-then-fine-tune pipeline. Without this control the observed gains cannot be attributed to the sleep data distribution rather than differences in training duration or hyperparameter effort, which is load-bearing for the central claim.

    Authors: We agree that matching the total optimization budget is essential to isolate the effect of the sleep pretraining distribution. In the original experiments, from-scratch models were trained for the same number of epochs as the fine-tuning stage, but the manuscript did not explicitly equate total steps to the combined pretraining plus fine-tuning budget. We have added a new ablation in the revised manuscript (Section 4) where scratch models receive an equivalent total step count; the performance gains from sleep pretraining persist under this control. revision: yes

  2. Referee: §5 (Results): No ablation is presented that trains scratch models for an extended number of steps matching the pretraining plus fine-tuning budget. Such an ablation is required to rule out optimization artifacts as the source of the reported improvements across the eight tasks.

    Authors: We acknowledge that this ablation is necessary to rule out optimization artifacts. We have performed the requested experiment and included the results in the revised Section 5. Across the eight downstream tasks, sleep-pretrained models continue to outperform scratch models trained for the matched extended budget, supporting that the improvements stem from the pretraining distribution rather than additional optimization steps. revision: yes

  3. Referee: Table 2 or equivalent results table: The manuscript supplies no statistical significance tests, confidence intervals, or error bars on the performance deltas, making it impossible to determine whether the claimed 'consistent improvements' are robust.

    Authors: We appreciate this feedback on statistical rigor. We have updated Table 2 in the revised manuscript to report means with standard deviations over five random seeds and added paired t-test p-values for the pretrain-then-fine-tune versus from-scratch comparisons. These additions confirm that the improvements are statistically significant on the majority of tasks. revision: yes

Circularity Check

0 steps flagged

No circularity in empirical sleep-to-non-sleep transfer claims

full rationale

The paper describes an empirical pipeline: multimodal contrastive pretraining exclusively on sleep polysomnography data, followed by evaluation on separate non-sleep EEG and ECG downstream tasks. Performance deltas are measured directly against scratch-trained baselines on the target datasets. No equations, procedures, or self-citations reduce the reported gains to quantities fitted on the evaluation tasks themselves. The pretraining distribution and evaluation distributions are disjoint, and the methodology contains no self-definitional steps, fitted-input predictions, or load-bearing uniqueness theorems imported from prior author work.

Axiom & Free-Parameter Ledger

0 free parameters · 0 axioms · 0 invented entities

The abstract provides no explicit free parameters, axioms, or invented entities. The central claim rests on the implicit domain assumption that sleep biosignals form a useful pretraining distribution for non-sleep EEG and ECG without severe negative transfer.

pith-pipeline@v0.9.0 · 5470 in / 1139 out tokens · 52732 ms · 2026-05-08T19:18:36.010551+00:00 · methodology

discussion (0)

Sign in with ORCID, Apple, or X to comment. Anyone can read and Pith papers without signing in.

Forward citations

Cited by 1 Pith paper

Reviewed papers in the Pith corpus that reference this work. Sorted by Pith novelty score.

  1. Mechanistic Interpretability of EEG Foundation Models via Sparse Autoencoders

    cs.LG 2026-05 unverdicted novelty 6.0

    Sparse autoencoders on EEG transformers identify three regimes of clinical concept encoding and reveal entanglements such as age-pathology confounding via a new steering selectivity metric.

Reference graph

Works this paper leans on

76 extracted references · 20 canonical work pages · cited by 1 Pith paper

  1. [1]

    The international federation of clinical neurophysiology , author=

    The ten-twenty electrode system of the international federation. The international federation of clinical neurophysiology , author=. Electroencephalogr. Clin. Neurophysiol. Suppl. , volume=

  2. [2]

    Circulation , volume=

    Practice standards for electrocardiographic monitoring in hospital settings: an American Heart Association scientific statement from the Councils on Cardiovascular Nursing, Clinical Cardiology, and Cardiovascular Disease in the Young: endorsed by the International Society of Computerized Electrocardiology and the American Association of Critical-Care Nurs...

    2004

  3. [3]

    International Classification of Sleep Disorders , howpublished =

  4. [4]

    arXiv preprint arXiv:2509.22810 , year=

    Introducing Multimodal Paradigm for Learning Sleep Staging PSG via General-Purpose Model , author=. arXiv preprint arXiv:2509.22810 , year=

    work page arXiv

  5. [5]

    arXiv preprint arXiv:2502.17481 , year=

    Toward Foundational Model for Sleep Analysis Using a Multimodal Hybrid Self-Supervised Learning Framework , author=. arXiv preprint arXiv:2502.17481 , year=

    work page arXiv

  6. [6]

    Lima, E. M. and Ribeiro, A. H. and Paix. Deep neural network-estimated electrocardiographic age as a mortality predictor , journal =. 2021 , volume =. doi:10.1038/s41467-021-25351-7 , pmid =

    work page doi:10.1038/s41467-021-25351-7 2021

  7. [7]

    and Fonseca, Pedro and van Meulen, Fokke B

    van Gorp, Hans and van Gilst, Merel M. and Fonseca, Pedro and van Meulen, Fokke B. and van Dijk, Johannes P. and Overeem, Sebastiaan and van Sloun, Ruud J. G. , year=. A Deep Generative Model for Five-Class Sleep Staging With Arbitrary Sensor Input , volume=. IEEE Journal of Biomedical and Health Informatics , publisher=. doi:10.1109/jbhi.2025.3565034 , number=

    work page doi:10.1109/jbhi.2025.3565034 2025

  8. [8]

    Journal of the American Medical Informatics Association , volume=

    The National Sleep Research Resource: towards a sleep data commons , author=. Journal of the American Medical Informatics Association , volume=. 2018 , publisher=

    2018

  9. [9]

    Racial/ethnic differences in sleep disturbances: the Multi-Ethnic Study of Atherosclerosis (

    Chen, Xiaoli and Wang, Rui and Zee, Phyllis and Lutsey, Pamela L and Javaheri, Sogol and Alc. Racial/ethnic differences in sleep disturbances: the Multi-Ethnic Study of Atherosclerosis (. Sleep , volume=. 2015 , publisher=

    2015

  10. [10]

    Journal of the American Geriatrics Society , volume=

    Associations between sleep architecture and sleep-disordered breathing and cognition in older community-dwelling men: the osteoporotic fractures in men sleep study , author=. Journal of the American Geriatrics Society , volume=. 2011 , publisher=

    2011

  11. [11]

    Psychiatry and Clinical Neurosciences , volume=

    Associations between self-reported parasomnias and psychiatric illness in 370,000 patients with sleep disorders , author=. Psychiatry and Clinical Neurosciences , volume=. 2024 , publisher=

    2024

  12. [12]

    Sleep , volume=

    The sleep heart health study: design, rationale, and methods , author=. Sleep , volume=. 1997 , publisher=

    1997

  13. [13]

    Advances in Neural Information Processing Systems , year =

    D. Advances in Neural Information Processing Systems , year =

  14. [14]

    Nature Medicine , pages=

    A multimodal sleep foundation model for disease prediction , author=. Nature Medicine , pages=. 2026 , publisher=

    2026

  15. [15]

    2024 , eprint=

    xLSTM: Extended Long Short-Term Memory , author=. 2024 , eprint=

    2024

  16. [16]

    2025 , eprint=

    Benchmarking ECG Foundational Models: A Reality Check Across Clinical Tasks , author=. 2025 , eprint=

    2025

  17. [17]

    npj digital Medicine , year =

    Maurer, Miriam Cindy and Hempel, Philip and Steinhaus, Kristin Elisabeth and Chereda, Hryhorii and Vollmer, Marcus and Krefting, Dagmar and Spicher, Nicolai and Hauschild, Anne-Christin , title =. npj digital Medicine , year =. doi:10.21203/rs.3.rs-7721630/v1 , url =

    work page doi:10.21203/rs.3.rs-7721630/v1

  18. [18]

    and Wang, C

    Yisimitila, T. and Wang, C. and Hou, M. and others , title =. npj Digital Medicine , volume =. 2026 , doi =

    2026

  19. [19]

    2022 , eprint=

    Advancing the State-of-the-Art for ECG Analysis through Structured State Space Models , author=. 2022 , eprint=

    2022

  20. [20]

    2025 , eprint=

    BenchECG and xECG: a benchmark and baseline for ECG foundation models , author=. 2025 , eprint=

    2025

  21. [21]

    Application of Graph Neural Networks on ECG data: A Systematic Literature Review , year=

    Müller, Alissa and Scheibl, Manuel and Uhe, Tobias and Schäbitz, Wolf-Rüdiger and Wrede, Britta , journal=. Application of Graph Neural Networks on ECG data: A Systematic Literature Review , year=

  22. [22]

    Nature communications , volume=

    Automatic diagnosis of the 12-lead ECG using a deep neural network , author=. Nature communications , volume=. 2020 , publisher=

    2020

  23. [23]

    Classification of 12-lead ECGs: the PhysioNet/Computing in Cardiology Challenge 2020.Physiological Measurement, 41(12):124003, December 2020

    Perez Alday, Erick A. and Gu, Annie and Shah, Amit J. and Robichaux, Chad and Wong, An-Kwok Ian and Liu, Chengyu and Liu, Feifei and Bahrami Rad, Ali and Elola, Andoni and Seyedi, Salman and Li, Qiao and Sharma, Ashish and Clifford, Gari D. and Reyna, Matthew A. , title =. Physiological Measurement , year =. doi:10.1088/1361-6579/abc960 , url =

    work page doi:10.1088/1361-6579/abc960

  24. [24]

    Deep Learning for ECG Analysis: Benchmarks and Insights from PTB-XL , year=

    Strodthoff, Nils and Wagner, Patrick and Schaeffter, Tobias and Samek, Wojciech , journal=. Deep Learning for ECG Analysis: Benchmarks and Insights from PTB-XL , year=

  25. [25]

    Sleep , volume=

    The 2007 AASM recommendations for EEG electrode placement in polysomnography: impact on sleep and cortical arousal scoring , author=. Sleep , volume=. 2011 , publisher=

    2007

  26. [26]

    Journal of Clinical Sleep Medicine , volume=

    The AASM recommended and acceptable EEG montages are comparable for the staging of sleep and scoring of EEG arousals , author=. Journal of Clinical Sleep Medicine , volume=. 2014 , publisher=

    2014

  27. [27]

    Neural Computation 9(8), 1735–1780 (1997)

    Hochreiter, Sepp and Schmidhuber, Jürgen , urldate =. Long Short-Term Memory , volume =. Neural Computation , shortjournal =. doi:10.1162/neco.1997.9.8.1735 , abstract =

    work page doi:10.1162/neco.1997.9.8.1735 1997

  28. [28]

    InceptionTime: Finding AlexNet for time series classification

    Ismail Fawaz, Hassan and Lucas, Benjamin and Forestier, Germain and Pelletier, Charlotte and Schmidt, Daniel F. and Weber, Jonathan and Webb, Geoffrey I. and Idoumghar, Lhassane and Muller, Pierre-Alain and Petitjean, François , urldate =. Data Mining and Knowledge Discovery , shortjournal =. doi:10.1007/s10618-020-00710-y , shorttitle =

    work page doi:10.1007/s10618-020-00710-y

  29. [29]

    In: IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR)

    He, Tong and Zhang, Zhi and Zhang, Hang and Zhang, Zhongyue and Xie, Junyuan and Li, Mu , booktitle =. 2019 , volume =. doi:10.1109/CVPR.2019.00065 , url =

    work page doi:10.1109/cvpr.2019.00065 2019

  30. [30]

    Time series classification from scratch with deep neural networks: A strong baseline , year=

    Wang, Zhiguang and Yan, Weizhong and Oates, Tim , booktitle=. Time series classification from scratch with deep neural networks: A strong baseline , year=

  31. [31]

    The accuracy of Apple Watch measurements: a living systematic review and meta-analysis , volume =

    Lambe, Rory and Baldwin, Maximus and O’Grady, Ben and Schumann, Moritz and Caulfield, Brian and Doherty, Cailbhe , urldate =. The accuracy of Apple Watch measurements: a living systematic review and meta-analysis , volume =. npj Digital Medicine , shortjournal =. doi:10.1038/s41746-025-02238-1 , shorttitle =

    work page doi:10.1038/s41746-025-02238-1

  32. [32]

    and Harmon, David M

    Attia, Zachi I. and Harmon, David M. and Dugan, Jennifer and Manka, Lukas and Lopez-Jimenez, Francisco and Lerman, Amir and Siontis, Konstantinos C. and Noseworthy, Peter A. and Yao, Xiaoxi and Klavetter, Eric W. and Halamka, John D. and Asirvatham, Samuel J. and Khan, Rita and Carter, Rickey E. and Leibovich, Bradley C. and Friedman, Paul A. , urldate =....

    work page doi:10.1038/s41591-022-02053-1

  33. [33]

    Hempel, Philip and Riedemann, Gabriel and Ribeiro, Antônio H. and Graf, Lennart and Sievers, Sören and Machreich, Till and Steinbrinker, Tabea and Schulmeister, Sophia and Haehling, Stephan Von and Krefting, Dagmar and Spicher, Nicolai , urldate =. Single-Lead, Single-Beat: Smartwatch. doi:10.36227/techrxiv.176281047.78677029/v1 , shorttitle =

    work page doi:10.36227/techrxiv.176281047.78677029/v1

  34. [34]

    The Twelfth International Conference on Learning Representations , year=

    Large Brain Model for Learning Generic Representations with Tremendous EEG Data in BCI , author=. The Twelfth International Conference on Learning Representations , year=

  35. [35]

    The Thirteenth International Conference on Learning Representations , year=

    CBraMod: A Criss-Cross Brain Foundation Model for EEG Decoding , author=. The Thirteenth International Conference on Learning Representations , year=

  36. [36]

    arXiv preprint arXiv:2407.20254 , year=

    EEGMamba: Bidirectional state space model with mixture of experts for EEG multi-task classification , author=. arXiv preprint arXiv:2407.20254 , year=

    work page arXiv

  37. [37]

    2024 IEEE International Symposium on Biomedical Imaging (ISBI) , pages=

    Neuro-gpt: Towards a foundation model for eeg , author=. 2024 IEEE International Symposium on Biomedical Imaging (ISBI) , pages=. 2024 , organization=

    2024

  38. [38]

    Frontiers in neuroscience , volume=

    The temple university hospital EEG data corpus , author=. Frontiers in neuroscience , volume=. 2016 , publisher=

    2016

  39. [39]

    2015 IEEE signal processing in medicine and biology symposium (SPMB) , pages=

    Automated identification of abnormal adult EEGs , author=. 2015 IEEE signal processing in medicine and biology symposium (SPMB) , pages=. 2015 , organization=

    2015

  40. [40]

    JAMIA open , volume=

    Ecg-fm: An open electrocardiogram foundation model , author=. JAMIA open , volume=. 2025 , publisher=

    2025

  41. [41]

    medRxiv , pages=

    HuBERT-ECG as a self-supervised foundation model for broad and scalable cardiac applications , author=. medRxiv , pages=. 2024 , publisher=

    2024

  42. [42]

    Plos one , volume=

    ECG classification using 1-D convolutional deep residual neural network , author=. Plos one , volume=. 2023 , publisher=

    2023

  43. [43]

    Advances in Neural Information Processing Systems , volume=

    Biot: Biosignal transformer for cross-data learning in the wild , author=. Advances in Neural Information Processing Systems , volume=

  44. [44]

    arXiv preprint arXiv:2512.09591 , year=

    Stanford Sleep Bench: Evaluating Polysomnography Pre-training Methods for Sleep Foundation Models , author=. arXiv preprint arXiv:2512.09591 , year=

    work page arXiv

  45. [45]

    International Conference on Machine Learning , pages=

    SleepFM: Multi-modal Representation Learning for Sleep Across Brain Activity, ECG and Respiratory Signals , author=. International Conference on Machine Learning , pages=. 2024 , organization=

    2024

  46. [46]

    Langley , title =

    P. Langley , title =. Proceedings of the 17th International Conference on Machine Learning (ICML 2000) , address =. 2000 , pages =

    2000

  47. [47]

    T. M. Mitchell. The Need for Biases in Learning Generalizations. 1980

    1980

  48. [48]

    M. J. Kearns , title =

  49. [49]

    Machine Learning: An Artificial Intelligence Approach, Vol. I. 1983

    1983

  50. [50]

    R. O. Duda and P. E. Hart and D. G. Stork. Pattern Classification. 2000

    2000

  51. [51]

    Suppressed for Anonymity , author=

  52. [52]

    Newell and P

    A. Newell and P. S. Rosenbloom. Mechanisms of Skill Acquisition and the Law of Practice. Cognitive Skills and Their Acquisition. 1981

    1981

  53. [53]

    A. L. Samuel. Some Studies in Machine Learning Using the Game of Checkers. IBM Journal of Research and Development. 1959

    1959

  54. [54]

    Speed: Scalable Preprocessing of EEG Data for Self-Supervised Learning , url=

    Gjølbye, Anders and Skerath, Lina and Lehn-Schiøler, William and Langer, Nicolas and Hansen, Lars Kai , year=. Speed: Scalable Preprocessing of EEG Data for Self-Supervised Learning , url=. doi:10.1109/mlsp58920.2024.10734749 , booktitle=

    work page doi:10.1109/mlsp58920.2024.10734749 2024

  55. [55]

    2025 , url =

    El Ouahidi, Yassine and Lys, Jonathan and Thölke, Philipp and Farrugia, Nicolas and Pasdeloup, Bastien and Gripon, Vincent and Jerbi, Karim and Lioi, Giulia , journal =. 2025 , url =

    2025

  56. [56]

    Neurology , volume=

    Development of Expert-Level Classification of Seizures and Rhythmic and Periodic Patterns During EEG Interpretation , author=. Neurology , volume=. 2023 , doi=

    2023

  57. [57]

    2025 , eprint=

    FEMBA: Efficient and Scalable EEG Analysis with a Bidirectional Mamba Foundation Model , author=. 2025 , eprint=

    2025

  58. [58]

    Frontiers in Human Neuroscience , year=

    BENDR: Using Transformers and a Contrastive Self-Supervised Learning Task to Learn From Massive Amounts of EEG Data , author=. Frontiers in Human Neuroscience , year=

  59. [59]

    Self-Supervised Electroencephalogram Representation Learning for Automatic Sleep Staging: Model Development and Evaluation Study

    Yang, Chaoqi and Xiao, Cao and Westover, M Brandon and Sun, Jimeng. Self-Supervised Electroencephalogram Representation Learning for Automatic Sleep Staging: Model Development and Evaluation Study. JMIR AI. 2023. doi:10.2196/46769

    work page doi:10.2196/46769 2023

  60. [60]

    2022 , eprint=

    Transformer Convolutional Neural Networks for Automated Artifact Detection in Scalp EEG , author=. 2022 , eprint=

    2022

  61. [61]

    Motor imagery EEG classification algorithm based on CNN-LSTM feature fusion network , journal =

    Hongli Li and Man Ding and Ronghua Zhang and Chunbo Xiu , keywords =. Motor imagery EEG classification algorithm based on CNN-LSTM feature fusion network , journal =. 2022 , issn =. doi:https://doi.org/10.1016/j.bspc.2021.103342 , url =

    work page doi:10.1016/j.bspc.2021.103342 2022

  62. [62]

    2025 , eprint=

    CBraMod: A Criss-Cross Brain Foundation Model for EEG Decoding , author=. 2025 , eprint=

    2025

  63. [63]

    2021 , eprint=

    Transformer-based Spatial-Temporal Feature Learning for EEG Decoding , author=. 2021 , eprint=

    2021

  64. [64]

    2023 , eprint=

    BrainBERT: Self-supervised representation learning for intracranial recordings , author=. 2023 , eprint=

    2023

  65. [65]

    2024 , eprint=

    EEGFormer: Towards Transferable and Interpretable Large-Scale EEG Foundation Model , author=. 2024 , eprint=

    2024

  66. [66]

    Biomedical Signal Processing and Control , volume=

    Deep learning for ECG classification: A comparative study of 1D and 2D representations and multimodal fusion approaches , author=. Biomedical Signal Processing and Control , volume=. 2024 , publisher=

    2024

  67. [67]

    2017 , address =

    Lopez, Silvia , title =. 2017 , address =

    2017

  68. [68]

    2015 IEEE Signal Processing in Medicine and Biology Symposium (SPMB) , pages=

    Improved EEG event classification using differential energy , author=. 2015 IEEE Signal Processing in Medicine and Biology Symposium (SPMB) , pages=. 2015 , organization=

    2015

  69. [69]

    Frontiers in neuroinformatics , volume=

    The temple university hospital seizure detection corpus , author=. Frontiers in neuroinformatics , volume=. 2018 , publisher=

    2018

  70. [70]

    CHB-MIT Scalp EEG Database.PhysioNet, June 2010

    Guttag, John , title =. 2010 , month = jun, note =. doi:10.13026/C2K01R , url =

    work page doi:10.13026/c2k01r 2010

  71. [71]

    , title =

    Shoeb, Ali H. , title =. 2009 , month =

    2009

  72. [72]

    and Amaral, Lu\'is A

    Goldberger, Ary L. and Amaral, Lu\'is A. N. and Glass, Leon and Hausdorff, Jeffrey M. and Ivanov, Plamen Ch. and Mark, Roger G. and Mietus, Joseph E. and Moody, George B. and Peng, Chung-Kang and Stanley, H. Eugene , title =. Circulation , volume =. 2000 , doi =

    2000

  73. [73]

    Epileptic Disorders , volume=

    Low-cost portable EEG device for bridging the diagnostic gap in resource-limited areas , author=. Epileptic Disorders , volume=. 2024 , publisher=

    2024

  74. [74]

    Strodthoff, Nils and others , title =

  75. [75]

    OpenECG: Benchmarking ECG Foundation Models with Public 1.2 Million Records.arXiv preprint arXiv:2503.00711, 2025

    OpenECG: Benchmarking ECG Foundation Models with Public 1.2 Million Records , author =. arXiv preprint arXiv:2503.00711 , year =

    work page arXiv

  76. [76]

    Cell Reports Medicine , year =

    Foundation model of ECG diagnosis: Diagnostics and prediction of cardiovascular diseases from electrocardiogram , author =. Cell Reports Medicine , year =. doi:10.1016/j.xcrm.2024.102033 , pmid =

    work page doi:10.1016/j.xcrm.2024.102033 2024