EEG and Brain-Computer Interfaces

This paper introduces a theory of stochastic spike timing and a method called SCA to accurately isolate neural sources in MEG/EEG data, outperforming PCA and ICA in separating overlapping brain signals.

This study improves MEG source inversion by using white matter surface locations and correspondence-based orientation estimation, showing better performance in evoked response fitting.

The MADE pipeline automates EEG preprocessing for pediatric data, addressing artifact contamination and short recording periods, and provides quality reports.

This study compares LCMV beamformer implementations across four MEG toolboxes, finding reliable source localization at typical SNR but differences in sensitivity to preprocessing parameters and SNR effects.

This study benchmarks EEG preprocessing methods by analyzing their impact on signal and event-related measures across multiple studies.

The paper optimizes the ADJUST algorithm for pediatric EEG data using geodesic nets, improving artifact classification and data retention compared to original ADJUST and ICLabel.

This paper introduces a method for accurately localizing on-scalp MEG sensors relative to the head, improving co-registration and source localization accuracy.

This paper improves neural activity reconstruction from low-density EEG using Multivariate Empirical Mode Decomposition (MEMD) and shows significant error reduction in both real and simulated data.

This paper introduces NID, a method to decompose cross-frequency coupled sources in EEG/MEG data, showing it reliably extracts nonlinear interactions even at low SNR and identifies alpha-beta phase-phase coupling in resting-state EEG.

The paper predicts bi-modal neurofeedback scores from EEG alone using a sparse regression model, showing improved correlation with combined fMRI-EEG sessions.

The paper introduces a method for visualizing mental state spaces to help users find effective BCI control strategies through interactive exploration.

This paper introduces grouped model reliance as a method to interpret neural decoding models, showing that alpha-band activity is consistently used but with inter-individual variability in frequency and topography.

This paper introduces a compact neural network with mixture z-scoring to automate sleep scoring, showing it avoids systematic bias present in standard methods.

This paper demonstrates reliable single-trial decoding of somatosensory imagery using fMRI, showing high accuracy in distinguishing left-foot and right-hand imagery and developing a real-time BCI communication system.

This paper introduces a method to record steady-state visual evoked potentials during real-world navigation and visual stimulation, finding reduced SSVEP amplitude during walking and preserved waveform sensitivity to visual hemifield.

This study compares MCMV and LCMV beamformers for estimating electrophysiological connectivity, showing MCMV reduces signal leakage and preserves short lag interactions.

The paper introduces a method to remove power line artifacts from electrophysiological data by combining spectral and spatial filtering, evaluated on synthetic and real data.

The paper introduces a method to identify spatially localized brain events using entropy mapping, achieving high sensitivity and specificity in predicting epilepsy foci.

This paper performs a large-scale EEG mega-analysis across 18 studies, showing that automated processing and consistent metadata enable channel- and source-level insights into spectral and amplitude characteristics.

This study evaluates digitizing methods for EEG electrodes, finding that more reliable methods reduce source estimation uncertainty, but all tested methods accurately identify Brodmann areas.

This paper optimizes ICA-based removal of ocular EEG artifacts in free viewing experiments by adjusting filtering and training parameters, achieving effective artifact removal with minimal neural activity distortion.

This paper presents a deep learning algorithm for computer-assisted EEG review in idiopathic generalized epilepsy, achieving high sensitivity and reducing human review time by 80%-99%.

The paper introduces unfold, a MATLAB toolbox for regression-based EEG analysis that handles overlapping signals and non-linear modeling.

This paper investigates using sequential probe words to decode semantic concepts from brain responses, finding no decrease in N400 signal strength or classification accuracy with later probes.

The paper introduces a framework to modify linear models using domain information to improve performance on noisy neuroimaging data.

This paper introduces a stimulus-aware spatial filtering method for improving neural response and temporal response function estimation in EEG data, showing improved performance over standard approaches.

This study proposes a zero-shot EEG-to-image decoding approach for image retrieval, achieving competitive accuracy across two EEG datasets using biologically inspired models.

This paper introduces a BCI system combining EEG2Code with deep learning, achieving high information transfer rates and accurate decoding of visual stimuli from EEG signals.

This paper performs a large-scale EEG analysis across 17 studies to identify cognitive aspects of event-related responses using HED tags and RSA, showing significant similarities in annotated features across diverse conditions.

This study investigates if alpha neurofeedback training with a low-cost EEG device improves alpha amplitudes and short-term memory, using a sham-controlled design.