{"status":"ok","message-type":"work","message-version":"1.0.0","message":{"indexed":{"date-parts":[[2026,9,25]],"date-time":"2026-09-25T21:35:01Z","timestamp":1790372101237,"version":"4.1.0"},"reference-count":24,"publisher":"MDPI AG","issue":"11","license":[{"start":{"date-parts":[[2026,5,23]],"date-time":"2026-05-23T00:00:00Z","timestamp":1779494400000},"content-version":"vor","delay-in-days":0,"URL":"https:\/\/creativecommons.org\/licenses\/by\/4.0\/"}],"funder":[{"name":"National Plan for NRRP Complementary Investments","award":["PNC0000003"],"award-info":[{"award-number":["PNC0000003"]}]}],"content-domain":{"domain":[],"crossmark-restriction":false},"short-container-title":["Sensors"],"abstract":"<jats:p>Brain\u2013computer interfaces (BCIs) based on motor imagery (MI) aim to convert electroencephalographic (EEG) activity into reliable device commands across users and recording setups. However, low signal-to-noise ratio and strong inter-subject variability still limit true \u201cplug-and-play\u201d deployment without lengthy calibration. To address these challenges, we propose a multi-branch convolutional long short-term memory (CNN-LSTM) architecture that jointly performs multi-scale temporal feature extraction and within-trial sequence modeling. The model employs four parallel 1D convolutional branches with distinct kernel sizes, each followed by an LSTM module and late fusion, combined with group normalization and supervision over sequences of sub-windows within each trial. We evaluate the approach on the EEG Motor Movement\/Imagery (EEGMMI) dataset from PhysioNet under strictly subject-independent conditions, and on the ISLab-MI Dataset, a 32-channel wearable-EEG collection designed to assess cross-setup robustness. On EEGMMI, the network achieves up to 82.63% accuracy for binary left\/right MI and 74.10% for a four-class task using 4 s trials under 5-fold cross-validation, outperforming an EEGNet-style baseline by 1\u201310% depending on class count and window length. Under a leave-one-subject-out protocol, the model attains 74.9% mean accuracy for a three-class MI task. Zero-shot transfer to ISLab-MI yields 64.60% and 63.02% accuracy in three- and four-class settings, respectively, while brief subject-specific fine-tuning using only 20% of each session improves performance to 81.38% and 73.48%. These findings show that combining multi-scale convolutional feature extraction with explicit sequence modeling and robust normalization yields accurate, data-efficient, and portable MI decoders suitable for practical BCI applications.<\/jats:p>","DOI":"10.3390\/s26113310","type":"journal-article","created":{"date-parts":[[2026,5,25]],"date-time":"2026-05-25T01:45:08Z","timestamp":1779673508000},"page":"3310","update-policy":"https:\/\/doi.org\/10.3390\/mdpi_crossmark_policy","source":"Crossref","is-referenced-by-count":1,"title":["A Robust Multi-Branch CNN-LSTM Architecture for Cross-Subject Motor Imagery Classification"],"prefix":"10.3390","volume":"26","author":[{"ORCID":"https:\/\/orcid.org\/0000-0002-8505-1581","authenticated-orcid":false,"given":"Simone","family":"Zini","sequence":"first","affiliation":[{"name":"Department of Informatics, Systems and Communication, University of Milano-Bicocca, Viale Sarca 336, 20126 Milano, Italy"}],"role":[{"vocabulary":"crossref","role":"author"}]},{"given":"Federico","family":"Bidone","sequence":"additional","affiliation":[{"name":"Department of Informatics, Systems and Communication, University of Milano-Bicocca, Viale Sarca 336, 20126 Milano, Italy"}],"role":[{"vocabulary":"crossref","role":"author"}]},{"ORCID":"https:\/\/orcid.org\/0000-0001-9112-0574","authenticated-orcid":false,"given":"Paolo","family":"Napoletano","sequence":"additional","affiliation":[{"name":"Department of Informatics, Systems and Communication, University of Milano-Bicocca, Viale Sarca 336, 20126 Milano, Italy"}],"role":[{"vocabulary":"crossref","role":"author"}]}],"member":"1968","published-online":{"date-parts":[[2026,5,23]]},"reference":[{"key":"ref_1","doi-asserted-by":"crossref","first-page":"767","DOI":"10.1016\/S1388-2457(02)00057-3","article-title":"Brain\u2013computer interfaces for communication and control","volume":"113","author":"Wolpaw","year":"2002","journal-title":"Clin. Neurophysiol."},{"key":"ref_2","doi-asserted-by":"crossref","unstructured":"Hossain, K.M., Islam, M.A., Hossain, S., Nijholt, A., and Ahad, M.A.R. (2022). Status of deep learning for EEG-based brain\u2013computer interface applications. Front. Comput. Neurosci., 16.","DOI":"10.3389\/fncom.2022.1006763"},{"key":"ref_3","doi-asserted-by":"crossref","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":"ref_4","doi-asserted-by":"crossref","unstructured":"Palumbo, A., Gramigna, V., Calabrese, B., and Ielpo, N. (2021). Motor-Imagery EEG-Based BCIs in Wheelchair Movement and Control: A Systematic Literature Review. Sensors, 21.","DOI":"10.36227\/techrxiv.14916537"},{"key":"ref_5","doi-asserted-by":"crossref","unstructured":"Zheng, Y., Wu, S., Chen, J., Yao, Q., and Zheng, S. (2025). Cross-Subject Motor Imagery Electroencephalogram Decoding with Domain Generalization. Bioengineering, 12.","DOI":"10.3390\/bioengineering12050495"},{"key":"ref_6","doi-asserted-by":"crossref","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":"ref_7","doi-asserted-by":"crossref","unstructured":"Zhang, J., Li, K., Yang, B., and Han, Y. (2023). Local and global convolutional transformer-based motor imagery EEG classification. Front. Neurosci., 17.","DOI":"10.3389\/fnins.2023.1219988"},{"key":"ref_8","doi-asserted-by":"crossref","unstructured":"Ding, Y., Robinson, N., Zeng, Q., Chen, D., Wai, A.A.P., Lee, T.S., and Guan, C. (2020). TSception: A deep learning framework for emotion detection using EEG. Proceedings of the 2020 International Joint Conference on Neural Networks (IJCNN), Glasgow, UK, 19\u201324 July 2020, IEEE.","DOI":"10.1109\/IJCNN48605.2020.9206750"},{"key":"ref_9","doi-asserted-by":"crossref","first-page":"016025","DOI":"10.1088\/1741-2552\/ab405f","article-title":"HS-CNN: A CNN with hybrid convolution scale for EEG motor imagery classification","volume":"17","author":"Dai","year":"2020","journal-title":"J. Neural Eng."},{"key":"ref_10","doi-asserted-by":"crossref","first-page":"710","DOI":"10.1109\/TNSRE.2022.3230250","article-title":"EEG conformer: Convolutional transformer for EEG decoding and visualization","volume":"31","author":"Song","year":"2022","journal-title":"IEEE Trans. Neural Syst. Rehabil. Eng."},{"key":"ref_11","doi-asserted-by":"crossref","unstructured":"Tao, Y., Sun, T., Muhamed, A., Genc, S., Jackson, D., Arsanjani, A., Yaddanapudi, S., Li, L., and Kumar, P. (2021). Gated transformer for decoding human brain EEG signals. Proceedings of the 2021 43rd Annual International Conference of the IEEE Engineering in Medicine & Biology Society (EMBC), Virtual, 1\u20135 November 2021, IEEE.","DOI":"10.1109\/EMBC46164.2021.9630210"},{"key":"ref_12","doi-asserted-by":"crossref","first-page":"2126","DOI":"10.1109\/TNSRE.2022.3194600","article-title":"A transformer-based approach combining deep learning network and spatial-temporal information for raw EEG classification","volume":"30","author":"Xie","year":"2022","journal-title":"IEEE Trans. Neural Syst. Rehabil. Eng."},{"key":"ref_13","doi-asserted-by":"crossref","unstructured":"Wang, X., Hersche, M., T\u00f6mekce, B., Kaya, B., Magno, M., and Benini, L. (2020). An accurate EEGNet-based motor-imagery brain\u2013computer interface for low-power edge computing. arXiv.","DOI":"10.1109\/MeMeA49120.2020.9137134"},{"key":"ref_14","doi-asserted-by":"crossref","first-page":"031005","DOI":"10.1088\/1741-2552\/aab2f2","article-title":"A review of classification algorithms for EEG-based brain\u2013computer interfaces: A 15-year update","volume":"15","author":"Lotte","year":"2018","journal-title":"J. Neural Eng."},{"key":"ref_15","doi-asserted-by":"crossref","first-page":"217987","DOI":"10.1155\/2011\/217987","article-title":"Multisubject learning for common spatial patterns in motor-imagery BCI","volume":"2011","author":"Devlaminck","year":"2011","journal-title":"Comput. Intell. Neurosci."},{"key":"ref_16","doi-asserted-by":"crossref","first-page":"1753","DOI":"10.1109\/TNSRE.2016.2627016","article-title":"Riemannian approaches in brain\u2013computer interfaces: A review","volume":"25","author":"Yger","year":"2017","journal-title":"IEEE Trans. Neural Syst. Rehabil. Eng."},{"key":"ref_17","doi-asserted-by":"crossref","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":"Hum. Brain Mapp."},{"key":"ref_18","doi-asserted-by":"crossref","first-page":"066004","DOI":"10.1088\/1741-2552\/ab3471","article-title":"A novel hybrid deep learning scheme for four-class motor imagery classification","volume":"16","author":"Zhang","year":"2019","journal-title":"J. Neural Eng."},{"key":"ref_19","doi-asserted-by":"crossref","first-page":"105288","DOI":"10.1016\/j.compbiomed.2022.105288","article-title":"A transfer learning-based CNN and LSTM hybrid deep learning model to classify motor imagery EEG signals","volume":"143","author":"Khademi","year":"2022","journal-title":"Comput. Biol. Med."},{"key":"ref_20","doi-asserted-by":"crossref","first-page":"158855","DOI":"10.1109\/ACCESS.2025.3605265","article-title":"SSTAF: Spatial-Spectral-Temporal Attention Fusion Transformer for Motor Imagery Classification","volume":"13","author":"Muna","year":"2025","journal-title":"IEEE Access"},{"key":"ref_21","doi-asserted-by":"crossref","unstructured":"Tang, X., and Zhang, X. (2020). Conditional adversarial domain adaptation neural network for motor imagery EEG decoding. Entropy, 22.","DOI":"10.3390\/e22010096"},{"key":"ref_22","doi-asserted-by":"crossref","first-page":"177","DOI":"10.1023\/A:1023437823106","article-title":"Mu and beta rhythm topographies during motor imagery and actual movements","volume":"12","author":"McFarland","year":"2000","journal-title":"Brain Topogr."},{"key":"ref_23","doi-asserted-by":"crossref","first-page":"e215","DOI":"10.1161\/01.CIR.101.23.e215","article-title":"PhysioBank, PhysioToolkit, and PhysioNet: Components of a new research resource for complex physiologic signals","volume":"101","author":"Goldberger","year":"2000","journal-title":"Circulation"},{"key":"ref_24","doi-asserted-by":"crossref","unstructured":"Ghimire, A., and Sekeroglu, K. (2022). Classification of EEG Motor Imagery Tasks Utilizing 2D Temporal Patterns with Deep Learning. Proceedings of the 14th International Joint Conference on Biomedical Engineering Systems and Technologies (BIOSTEC), SciTePress.","DOI":"10.5220\/0011069400003209"}],"container-title":["Sensors"],"original-title":[],"language":"en","link":[{"URL":"https:\/\/www.mdpi.com\/1424-8220\/26\/11\/3310\/pdf","content-type":"unspecified","content-version":"vor","intended-application":"similarity-checking"}],"deposited":{"date-parts":[[2026,5,25]],"date-time":"2026-05-25T01:52:11Z","timestamp":1779673931000},"score":1,"resource":{"primary":{"URL":"https:\/\/www.mdpi.com\/1424-8220\/26\/11\/3310"}},"subtitle":[],"short-title":[],"issued":{"date-parts":[[2026,5,23]]},"references-count":24,"journal-issue":{"issue":"11","published-online":{"date-parts":[[2026,6]]}},"alternative-id":["s26113310"],"URL":"https:\/\/doi.org\/10.3390\/s26113310","relation":{},"ISSN":["1424-8220"],"issn-type":[{"value":"1424-8220","type":"electronic"}],"subject":[],"published":{"date-parts":[[2026,5,23]]}}}