{"status":"ok","message-type":"work","message-version":"1.0.0","message":{"indexed":{"date-parts":[[2026,9,28]],"date-time":"2026-09-28T13:56:37Z","timestamp":1790603797895,"version":"4.1.0"},"reference-count":56,"publisher":"Frontiers Media SA","license":[{"start":{"date-parts":[[2026,9,28]],"date-time":"2026-09-28T00:00:00Z","timestamp":1790553600000},"content-version":"vor","delay-in-days":0,"URL":"https:\/\/creativecommons.org\/licenses\/by\/4.0\/"}],"funder":[{"DOI":"10.13039\/501100002915","name":"Fondation pour la Recherche M\u00e9dicale","doi-asserted-by":"publisher","id":[{"id":"10.13039\/501100002915","id-type":"DOI","asserted-by":"publisher"}]}],"content-domain":{"domain":["frontiersin.org"],"crossmark-restriction":true},"short-container-title":["Front. Neuroergonomics"],"abstract":"<jats:sec>\n                    <jats:title>Objective<\/jats:title>\n                    <jats:p>Deep learning classification methods are increasingly popular in the brain-computer interface (BCI) field due to their ability to learn complex, non-linear representations from EEG data. Unlike traditional machine learning approaches, which typically rely on hand-crafted features, deep learning models autonomously extract spatio-temporal features directly from the raw or minimally preprocessed signal. However, this flexibility may come at the cost of reduced interpretability and increased susceptibility to confounding factors. This study aims to investigate whether the inclusion of task-related visual cues in the input EEG window biases deep learning models by leading them to exploit irrelevant features, thereby compromising their neurophysiological validity.<\/jats:p>\n                  <\/jats:sec>\n                  <jats:sec>\n                    <jats:title>Approach<\/jats:title>\n                    <jats:p>This study investigates how including EEG signals measured during visual cue presentation on classification accuracy and feature extraction in motor imagery (MI)-based BCIs. Three classification pipelines were evaluated: CSP+LDA, a traditional machine learning approach; DeepConvNet, a convolutional neural network for EEG classification; and REVE, a Transformer-based EEG foundation model. All methods are trained and tested on two large scale open-access EEG datasets using two epoching strategies, with or without including the visual cue onset, to assess their impact on model performance and neurophysiological relevance. Attribution maps (e.g., Saliency maps) and CSP patterns are used to interpret the features learned by each model.<\/jats:p>\n                  <\/jats:sec>\n                  <jats:sec>\n                    <jats:title>Main results<\/jats:title>\n                    <jats:p>Including the visual cue onset significantly improved the classification accuracy of DeepConvNet and REVE. However, attribution analysis showed that the model mainly relied on early parietal activations, most likely cue-related visual evoked potentials, rather than MI-related sensorimotor rhythms (ERD\/ERS), which were more prominent when the cue onset was excluded. In contrast, CSP+LDA showed minimal accuracy changes across conditions, with consistent motor cortex activation patterns.<\/jats:p>\n                  <\/jats:sec>\n                  <jats:sec>\n                    <jats:title>Significance<\/jats:title>\n                    <jats:p>These findings demonstrate that higher classification performance does not necessarily imply neurophysiological validity. In particular, deep learning models may exploit confounding signals, such as visual cue processing, which undermines their interpretability and applicability in asynchronous or real-world BCIs, in which there is no such cue. We advocate for systematically applying explainability tools in BCI research to ensure classifiers rely on task-relevant brain activity. Additionally, we emphasize the importance of excluding visual cue periods from EEG segments containing reposenses evoked by the visual cue onset from the input to deep learning models.<\/jats:p>\n                  <\/jats:sec>","DOI":"10.3389\/fnrgo.2026.1804143","type":"journal-article","created":{"date-parts":[[2026,9,28]],"date-time":"2026-09-28T13:11:42Z","timestamp":1790601102000},"update-policy":"https:\/\/doi.org\/10.3389\/crossmark-policy","source":"Crossref","is-referenced-by-count":0,"title":["Visual cues in MI-BCI induce biases in EEG classification models with deep learning but not with standard machine learning"],"prefix":"10.3389","volume":"7","author":[{"given":"David","family":"Trocellier","sequence":"first","affiliation":[{"name":"Inria Center at the University of Bordeaux","place":["Talence, France"]},{"name":"LaBRI","place":["Talence, France"]}],"role":[{"vocabulary":"crossref","role":"author"}]},{"given":"Simon","family":"Kojima","sequence":"additional","affiliation":[{"name":"Inria Center at the University of Bordeaux","place":["Talence, France"]}],"role":[{"vocabulary":"crossref","role":"author"}]},{"given":"Bernard","family":"N'kaoua","sequence":"additional","affiliation":[{"name":"BPH, Bordeaux Population Health, University of Bordeaux","place":["Bordeaux, France"]}],"role":[{"vocabulary":"crossref","role":"author"}]},{"given":"Fabien","family":"Lotte","sequence":"additional","affiliation":[{"name":"Inria Center at the University of Bordeaux","place":["Talence, France"]},{"name":"LaBRI","place":["Talence, France"]}],"role":[{"vocabulary":"crossref","role":"author"}]}],"member":"1965","published-online":{"date-parts":[[2026,9,28]]},"reference":[{"key":"B1","doi-asserted-by":"publisher","first-page":"127271","DOI":"10.1109\/ACCESS.2023.3329678","article-title":"Deep learning in EEG-based BCIs: a comprehensive review of transformer models, advantages, challenges, and applications","volume":"11","author":"Abibullaev","year":"2023","journal-title":"IEEE Access"},{"key":"B2","doi-asserted-by":"publisher","first-page":"14681","DOI":"10.1007\/s00521-021-06352-5","volume":"35","author":"Altaheri","year":"2023"},{"key":"B3","doi-asserted-by":"publisher","DOI":"10.1088\/1741-2552\/ac74e0","article-title":"How to successfully classify EEG in motor imagery BCI: a metrological analysis of the state of the art","author":"Arpaia","year":"2022","journal-title":"J. Neural Eng"},{"key":"B4","doi-asserted-by":"publisher","first-page":"289","DOI":"10.1111\/j.2517-6161.1995.tb02031.x","article-title":"Controlling the false discovery rate: a practical and powerful approach to multiple testing","volume":"57","author":"Benjamini","year":"1995","journal-title":"J. Royal Statist. Soc.: Series B (Methodological)"},{"key":"B5","doi-asserted-by":"publisher","first-page":"814","DOI":"10.1016\/j.neuroimage.2010.06.048","article-title":"Single-trial analysis and classification of ERP components\u2014A tutorial","volume":"56","author":"Blankertz","year":"2011","journal-title":"NeuroImage"},{"key":"B6","doi-asserted-by":"publisher","first-page":"41","DOI":"10.1109\/MSP.2008.4408441","article-title":"Optimizing spatial filters for robust EEG single-trial analysis","volume":"25","author":"Blankertz","year":"2008","journal-title":"IEEE Signal Proc. Magaz"},{"key":"B7","doi-asserted-by":"publisher","first-page":"136","DOI":"10.3389\/fnins.2015.00136","article-title":"Using neurophysiological signals that reflect cognitive or affective state: six recommendations to avoid common pitfalls","volume":"9","author":"Brouwer","year":"2015","journal-title":"Front. Neurosci"},{"key":"B8","doi-asserted-by":"publisher","first-page":"021002","DOI":"10.1088\/1741-2552\/adbfbd","article-title":"The NERVE-ML (neural engineering reproducibility and validity essentials for machine learning) checklist: ensuring machine learning advances neural engineering","volume":"22","author":"Carlson","year":"2025","journal-title":"J. Neural Eng"},{"key":"B9","author":"Chevallier","year":"2024","journal-title":"The Largest EEG-Based BCI Reproducibility Study For Open Science: the MOABB Benchmark"},{"key":"B10","doi-asserted-by":"publisher","first-page":"580","DOI":"10.1038\/s41597-023-02445-z","article-title":"A large EEG database with users' profile information for motor imagery brain-computer interface research","volume":"10","author":"Dreyer","year":"2023","journal-title":"Sci Data"},{"key":"B11","doi-asserted-by":"publisher","first-page":"044002","DOI":"10.1088\/1741-2552\/ad6793","article-title":"Benchmarking brain\u2013computer interface algorithms: Riemannian approaches vs convolutional neural networks","volume":"21","author":"Eder","year":"2024","journal-title":"J. Neural Eng"},{"key":"B12","doi-asserted-by":"publisher","DOI":"10.52202\/085713-0760","article-title":"REVE: A foundation model for EEG: Adapting to any setup with large-scale pretraining on 25,000 subjects","author":"El Ouahidi","year":"2025","journal-title":"Adv. Neural Inform. Proc. Syst"},{"key":"B13","doi-asserted-by":"publisher","first-page":"480","DOI":"10.1016\/j.clinph.2006.10.019","article-title":"EMG and EOG artifacts in brain computer interface systems: a survey","volume":"118","author":"Fatourechi","year":"2007","journal-title":"Clinical Neurophysiol"},{"key":"B14","doi-asserted-by":"publisher","first-page":"179","DOI":"10.1111\/j.1469-1809.1936.tb02137.x","article-title":"The use of multiple measurements in taxonomic problems","volume":"7","author":"Fisher","year":"1936","journal-title":"Ann. Eugenics"},{"key":"B15","doi-asserted-by":"crossref","first-page":"65","DOI":"10.1007\/978-3-030-60460-8_6","article-title":"\u201cFinal results of multi-center randomized controlled trials of BCI-controlled hand exoskeleton complex assisting post-stroke motor function recovery,\u201d","author":"Frolov","year":"2021","journal-title":"Brain-Computer Interface Research: A State-of-the-Art Summary 9, SpringerBriefs in Electrical and Computer Engineering"},{"key":"B16","first-page":"1","article-title":"\u201cBrain\u2013computer interfaces: a gentle introduction,\u201d","author":"Graimann","year":"2009","journal-title":"Brain-Computer Interfaces"},{"key":"B17","unstructured":"\u201cHierarchical internal representation of spectral features in deep convolutional networks trained for EEG decoding,\u201d\n          \n          \n            \n              Hartmann\n              K. G.\n            \n            \n              Schirrmeister\n              R. T.\n            \n            \n              Ball\n              T.\n            \n          \n          2018 6th International Conference on Brain-Computer Interface (BCI)\n          \n          2018"},{"key":"B18","doi-asserted-by":"publisher","first-page":"793","DOI":"10.1007\/978-3-030-01219-9_47","article-title":"\u201cWomen also snowboard: Overcoming bias in captioning models,\u201d","author":"Hendricks","year":"2018","journal-title":"Computer Vision - ECCV 2018, Vol. 11207"},{"key":"B19","doi-asserted-by":"publisher","first-page":"781","DOI":"10.1073\/pnas.95.3.781","article-title":"Event-related brain potentials in the study of visual selective attention","volume":"95","author":"Hillyard","year":"1998","journal-title":"Proc. Natl. Acad. Sci. U.S.A"},{"key":"B20","doi-asserted-by":"publisher","first-page":"3","DOI":"10.36518\/2689-0216.1188","article-title":"What external variables affect Sensorimotor Rhythm Brain-Computer Interface (SMR-BCI) performance?","volume":"2","author":"Horowitz","year":"2021","journal-title":"HCA Healthcare J. Med"},{"key":"B21","article-title":"\u201cOptimizing visual cues for brain-computer interfaces,\u201d","author":"Jacinto","year":"2009","journal-title":"World Congress on Medical Physics and Biomedical Engineering"},{"key":"B22","doi-asserted-by":"publisher","first-page":"583","DOI":"10.1038\/s41586-021-03819-2","article-title":"Highly accurate protein structure prediction with AlphaFold","volume":"596","author":"Jumper","year":"2021","journal-title":"Nature"},{"key":"B23","doi-asserted-by":"publisher","first-page":"1","DOI":"10.1145\/3505244","article-title":"Transformers in vision: a survey","volume":"200","author":"Khan","year":"2022","journal-title":"ACM Comp. Surv"},{"key":"B24","doi-asserted-by":"publisher","first-page":"15","DOI":"10.1016\/j.expneurol.2012.09.014","article-title":"The ups and downs of beta oscillations in sensorimotor cortex","volume":"245","author":"Kilavik","year":"2013","journal-title":"Exp. Neurol"},{"key":"B25","doi-asserted-by":"publisher","first-page":"021001","DOI":"10.1088\/1741-2552\/ae4455","article-title":"EEG foundation models: a critical review of current progress and future directions","volume":"23","author":"Kuruppu","year":"2026","journal-title":"J. Neural Eng"},{"key":"B26","doi-asserted-by":"publisher","first-page":"3839","DOI":"10.1109\/TNNLS.2019.2946869","article-title":"Subject-independent brain\u2013computer interfaces based on deep convolutional neural networks","volume":"31","author":"Kwon","year":"2019","journal-title":"IEEE Trans. Neural Netw. Learn. Syst"},{"key":"B27","doi-asserted-by":"publisher","first-page":"056013","DOI":"10.1088\/1741-2552\/aace8c","article-title":"EEGNet: A compact convolutional neural network for EEG-based brain\u2013computer interfaces","volume":"15","author":"Lawhern","year":"2018","journal-title":"J. Neural Eng"},{"key":"B28","doi-asserted-by":"publisher","first-page":"giz002","DOI":"10.1093\/gigascience\/giz002","article-title":"EEG dataset and OpenBMI toolbox for three BCI paradigms: an investigation into BCI illiteracy","volume":"8","author":"Lee","year":"2019","journal-title":"GigaScience"},{"key":"B29","author":"Lotte","year":"2014","journal-title":"A Tutorial on EEG Signal Processing Techniques for Mental State Recognition in Brain-Computer Interfaces"},{"key":"B30","doi-asserted-by":"publisher","first-page":"031005","DOI":"10.1088\/1741-2552\/aab2f2","article-title":"A review of classification algorithms for EEG-based brain-computer interfaces: a 10 year update","volume":"15","author":"Lotte","year":"2018","journal-title":"J. Neural Eng"},{"key":"B31","doi-asserted-by":"publisher","DOI":"10.48550\/arXiv.2104.01233","article-title":"FBCNet: A multi-view convolutional neural network for brain-computer interface","author":"Mane","year":"2021","journal-title":"arXiv"},{"key":"B32","doi-asserted-by":"publisher","first-page":"1","DOI":"10.1016\/j.dsp.2017.10.011","article-title":"Methods for interpreting and understanding deep neural networks","volume":"73","author":"Montavon","year":"2018","journal-title":"Digital Signal Proc"},{"key":"B33","doi-asserted-by":"publisher","first-page":"233","DOI":"10.1080\/1086508X.2008.11079688","article-title":"Visual evoked potentials","volume":"48","author":"Nehamkin","year":"2008","journal-title":"Am. J. Electroneurodiagnostic Technol"},{"key":"B34","doi-asserted-by":"publisher","first-page":"1681","DOI":"10.1016\/j.clinph.2005.03.013","article-title":"Long-term stability and consistency of EEG event-related (de-)synchronization across different cognitive tasks","volume":"116","author":"Neuper","year":"2005","journal-title":"Clini. Neurophysiol"},{"key":"B35","doi-asserted-by":"publisher","first-page":"73","DOI":"10.1109\/MSP.2022.3142719","article-title":"Robust explainability: a tutorial on gradient-based attribution methods for deep neural networks","volume":"39","author":"Nielsen","year":"2022","journal-title":"IEEE Signal Proc. Magazi"},{"key":"B36","doi-asserted-by":"publisher","first-page":"1293","DOI":"10.1523\/JNEUROSCI.5364-11.2012","article-title":"Human motor plasticity induced by mirror visual feedback","volume":"32","author":"Nojima","year":"2012","journal-title":"J. Neurosci"},{"key":"B37","doi-asserted-by":"publisher","first-page":"1842","DOI":"10.1016\/S1388-2457(99)00141-8","article-title":"Event-related EEG\/MEG synchronization and desynchronization: basic principles","volume":"110","author":"Pfurtscheller","year":"1999","journal-title":"Clin. Neurophysiol"},{"key":"B38","doi-asserted-by":"publisher","first-page":"1123","DOI":"10.1109\/5.939829","article-title":"Motor imagery and direct brain-computer communication","volume":"89","author":"Pfurtscheller","year":"2001","journal-title":"Proc. IEEE"},{"key":"B39","doi-asserted-by":"publisher","first-page":"379","DOI":"10.3389\/fnhum.2013.00379","article-title":"Brisk heart rate and EEG changes during execution and withholding of cue-paced foot motor imagery","volume":"7","author":"Pfurtscheller","year":"2013","journal-title":"Front. Human Neurosci"},{"key":"B40","doi-asserted-by":"publisher","first-page":"24","DOI":"10.1016\/j.clinph.2008.09.027","article-title":"Could the beta rebound in the EEG be suitable to realize a \u201cbrain switch\u201d?","volume":"120","author":"Pfurtscheller","year":"2009","journal-title":"Clin. Neurophysiol"},{"key":"B41","doi-asserted-by":"publisher","first-page":"441","DOI":"10.1109\/86.895946","article-title":"Optimal spatial filtering of single trial EEG during imagined hand movement","volume":"8","author":"Ramoser","year":"2000","journal-title":"IEEE Trans. Rehabilitat. Eng"},{"key":"B42","doi-asserted-by":"crossref","first-page":"1003","DOI":"10.1109\/ICORR.2007.4428546","article-title":"\u201cControlling a wheelchair using a BCI with low information transfer rate,\u201d","volume-title":"2007 IEEE 10th International Conference on Rehabilitation Robotics","author":"Rebsamen","year":"2007"},{"key":"B43","doi-asserted-by":"crossref","unstructured":"9\n          11\n          \n            \n              Riascos\n              J.\n            \n            \n              Molinas\n              M.\n            \n            \n              Lotte\n              F.\n            \n          \n          Machine Learning Methods for BCI: Challenges, Pitfalls and Promises\n          \n          2024","DOI":"10.14428\/esann\/2024.ES2024-4"},{"key":"B44","author":"Ryu","year":"2018","journal-title":"InclusiveFaceNet: Improving Face Attribute Detection with Race and Gender Diversity"},{"key":"B45","doi-asserted-by":"publisher","first-page":"1043","DOI":"10.1016\/j.neuroimage.2005.06.059","article-title":"Neuronal spatiotemporal pattern discrimination: the dynamical evolution of seizures","volume":"28","author":"Schiff","year":"2005","journal-title":"NeuroImage"},{"key":"B46","doi-asserted-by":"publisher","first-page":"5391","DOI":"10.1002\/hbm.23730","article-title":"Deep learning with convolutional neural networks for EEG decoding and visualization","volume":"38","author":"Schirrmeister","year":"2017","journal-title":"Human Brain Mapping"},{"key":"B47","doi-asserted-by":"publisher","first-page":"255","DOI":"10.1007\/s11831-024-10134-2","article-title":"A comprehensive review of bias in deep learning models: methods, impacts, and future directions","volume":"32","author":"Shah","year":"2025","journal-title":"Arch. Comput. Methods Eng"},{"key":"B48","first-page":"3145","article-title":"\u201cLearning important features through propagating activation differences,\u201d","volume-title":"Proceedings of the 34th International Conference on Machine Learning","author":"Shrikumar","year":"2017"},{"key":"B49","author":"Simonyan","year":"2014","journal-title":"Deep Inside Convolutional Networks: Visualising Image Classification Models and Saliency Maps"},{"key":"B50","doi-asserted-by":"publisher","first-page":"792","DOI":"10.1038\/s41551-025-01414-8","article-title":"A brain-computer interface working definition","volume":"9","author":"Slutzky","year":"2025","journal-title":"Nat. Biomed. Eng"},{"key":"B51","first-page":"3319","article-title":"\u201cAxiomatic attribution for deep networks,\u201d","volume-title":"Proceedings of the 34th International Conference on Machine Learning","author":"Sundararajan","year":"2017"},{"key":"B52","doi-asserted-by":"publisher","first-page":"120253","DOI":"10.1016\/j.neuroimage.2023.120253","article-title":"Class imbalance should not throw you off balance: Choosing the right classifiers and performance metrics for brain decoding with imbalanced data","volume":"277","author":"Th\u00f6lke","year":"2023","journal-title":"NeuroImage"},{"key":"B53","doi-asserted-by":"publisher","first-page":"e0268880","DOI":"10.1371\/journal.pone.0268880","article-title":"Classification of motor imagery EEG using deep learning increases performance in inefficient BCI users","volume":"17","author":"Tibrewal","year":"2022","journal-title":"PLoS ONE"},{"key":"B54","doi-asserted-by":"publisher","first-page":"4793","DOI":"10.1109\/TNNLS.2020.3027314","article-title":"A survey on Explainable Artificial Intelligence (XAI): toward medical XAI","volume":"32","author":"Tjoa","year":"2021","journal-title":"IEEE Trans. Neural Netw. Learn. Syst"},{"key":"B55","article-title":"\u201cAttention is all you need,\u201d","author":"Vaswani","year":"2017","journal-title":"Advances in Neural Information Processing Systems"},{"key":"B56","first-page":"205","article-title":"\u201c2021 BEETL competition: advancing transfer learning for subject independence and heterogenous EEG data sets,\u201d","volume-title":"Proceedings of the NeurIPS 2021 Competitions and Demonstrations Track","author":"Wei","year":"2022"}],"container-title":["Frontiers in Neuroergonomics"],"original-title":[],"link":[{"URL":"https:\/\/www.frontiersin.org\/articles\/10.3389\/fnrgo.2026.1804143\/full","content-type":"unspecified","content-version":"vor","intended-application":"similarity-checking"}],"deposited":{"date-parts":[[2026,9,28]],"date-time":"2026-09-28T13:11:46Z","timestamp":1790601106000},"score":1,"resource":{"primary":{"URL":"https:\/\/www.frontiersin.org\/articles\/10.3389\/fnrgo.2026.1804143\/full"}},"subtitle":[],"short-title":[],"issued":{"date-parts":[[2026,9,28]]},"references-count":56,"alternative-id":["10.3389\/fnrgo.2026.1804143"],"URL":"https:\/\/doi.org\/10.3389\/fnrgo.2026.1804143","relation":{},"ISSN":["2673-6195"],"issn-type":[{"value":"2673-6195","type":"electronic"}],"subject":[],"published":{"date-parts":[[2026,9,28]]},"article-number":"1804143"}}