Neighbourhood-clipped latent space of VAEs with spatially preserved EEG topographic maps for ocular artefact reduction

Simple item page
dc.contributor.authorAhmed, Taufique
dc.contributor.authorLongo, Luca
dc.contributor.editorLombardi, Angela
dc.contributor.editorBrattico, Elvira
dc.contributor.editorWang, Shuqiang
dc.contributor.editorKuai, Hongzhi
dc.date.accessioned2026-08-11T14:00:03Z
dc.date.available2026-08-11T14:00:03Z
dc.date.embargoedUntil2027-07-02
dc.date.issued2026-07-02
dc.description© 2026, the Author(s), under exclusive license to Springer Nature Singapore Pte Ltd.
dc.description.abstractElectroencephalographic (EEG) recordings are often affected by artefacts such as eye blinks, which complicate their analysis. Although multiple techniques exist to detect and remove artefacts, many require manual intervention. This study presents a novel, self-supervised, and fully automated approach to identify and reduce artefacts in EEG signals using a Variational Autoencoder (VAE) framework. In this approach, subject-specific VAEs with convolutional layers are trained on spatially preserved EEG topographic maps. An anomaly detection strategy based on the negative log-likelihood of activated latent vectors from the training data is employed to identify abnormal topomaps, assigning each input an anomaly score. Input topomaps exceeding a defined threshold, together with their neighbouring topomaps, are clipped using a standard IQR-based method, combining clipping and mitigation to influence surrounding regions affected by eye blinks. The reconstructed EEG signals are then compared against a baseline created using an offline Independent Component Analysis (ICA) method with automated detection of artefact components inspired by the FASTER methodology. Results indicate improved signal-to-noise ratio (SNR) and peak signal-to-noise ratio (PSNR) in channels such as FP1 and FP2, while other channels show comparable performance to ICA Fast. Additionally, mean absolute error (MAE), normalised root mean square error (NRMSE), and correlation coefficients demonstrate that the reconstructed signals maintain comparable quality to the baseline. The findings further show that the method preserves non-artifactual segments, maintaining their neural dynamics. Overall, this study introduces a fully automated, subject-specific approach for EEG artefact identification and denoising using latent space neighbourhood clipping of the latent space representation of anomalous topomaps.en
dc.format.extent12
dc.format.extent2166389
dc.identifier.authororcidAhmed, Taufique
dc.identifier.authororcidLongo, Luca§0000-0002-2718-5426
dc.identifier.authororcidLombardi, Angela
dc.identifier.authororcidBrattico, Elvira
dc.identifier.authororcidWang, Shuqiang
dc.identifier.authororcidKuai, Hongzhi
dc.identifier.citationAhmed, T. and Longo, L. (2026) 'Neighbourhood-clipped latent space of VAEs with spatially preserved EEG topographic maps for ocular artefact reduction', in Lombardi, A., Brattico, E., Wang. S. and Kuai, H. (eds) 18th International Conference on Brain Informatics (BI'25). Lecture Notes in Computer Science, 16347 LNAI, Springer Science and Business Media Deutschland GmbH, pp. 232-243. https://doi.org/10.1007/978-981-95-9575-4_18en
dc.identifier.citationconference
dc.identifier.doi10.1007/978-981-95-9575-4_18
dc.identifier.endpage243
dc.identifier.isbn9789819595747
dc.identifier.issn0302-9743
dc.identifier.journaltitle18th International Conference on Brain Informatics (BI'25)
dc.identifier.journaltitle18th International Conference on Brain Informatics, BI 2025
dc.identifier.otherORCID: /0000-0002-2718-5426/work/223399170
dc.identifier.startpage232
dc.identifier.urihttps://hdl.handle.net/10468/19108
dc.identifier.urlhttps://www.scopus.com/pages/publications/105045217985
dc.language.isoeng
dc.publisherSpringer Science and Business Media Deutschland GmbH
dc.relation.ispartofseriesLecture Notes in Computer Science
dc.subjectArtefacts removal
dc.subjectDeep learning
dc.subjectElectroencephalography
dc.subjectExplainable AI
dc.subjectFull automation
dc.subjectInterpretability
dc.subjectLatent space
dc.subjectSpectral topographic maps
dc.subjectSubject-specific
dc.subjectVariational autoencoder
dc.subject[ComputerScience]
dc.subjectTheoretical Computer Science
dc.subjectGeneral Computer Science
dc.titleNeighbourhood-clipped latent space of VAEs with spatially preserved EEG topographic maps for ocular artefact reductionen
dc.typeConference contribution (Peer reviewed)
Files
Original bundle
Now showing 1 - 1 of 1
Thumbnail Image
Name:
Taufique_Brain_informatics_Italy-1.pdf
Size:
2.07 MB
Format:
Adobe Portable Document Format
Collections
School of Computer Science and Information Technology