EEG and Brain-Computer Interfaces

This paper reanalyzed EEG data using three software tools to assess how preprocessing variability affects results.

This paper introduces FLEX-MUSIC, a method for localizing neural sources in M/EEG data that also estimates source extent, achieving high accuracy and efficiency.

This paper validates MEG source imaging by comparing it to an intracranial EEG atlas, finding that MEG accurately estimates oscillatory patterns in lateral regions but underestimates deep regions.

The paper introduces a semi-automatic artifact removal tool for high-density sleep EEG data, achieving high restoration rates of bad epochs through channel-wise interpolation.

EEG-LLAMAS improves real-time EEG-fMRI neurofeedback by reducing BCG artifacts with low latency, enabling closed-loop experiments previously difficult to perform.

This paper provides a protocol for setting up and recording high-quality ear-EEG data using cEEGrids for extended periods in and beyond the lab.

This paper trains study-independent mind-wandering classifiers from EEG using CNNs, finding limited generalizability but improved performance with single-trial ERP features.

This paper develops CNN encoders for parametric t-SNE to analyze EEG data, producing low-dimensional representations that capture EEG characteristics and generalize to new data.

The paper introduces an open-source deep learning classifier for sleep staging that achieves high accuracy and is easy to use across various electrode setups.

RELAX is an automated EEG preprocessing pipeline that effectively reduces artifacts while preserving neural signals, outperforming existing methods in blink and muscle artifact removal.

This paper evaluates the RELAX pipeline for cleaning EEG data before ERP analysis, finding it effective and reliable for artifact reduction and ERP estimation.

The paper evaluates EEG preprocessing methods and finds that most automated corrections do not improve ERP significance, with only one pipeline performing better than simple high-pass filtering.

This study investigates whether long-range alpha-synchronisation can serve as a control signal for BCI by analyzing EEG data during a covert attention task, finding it unreliable at the single-trial level.

The paper assesses ERP data quality across paradigms, participants, and scoring procedures using the SME metric, finding large variations in quality and highlighting the impact of quantification algorithms.

The paper analyzes intracranial recordings to compare speech activity detection across overt, mouthed, and imagined speech, finding a nested hierarchy of relevant brain channels.

This study shows low-cost open-source EEG amplifiers can detect stimulus-induced gamma oscillations in humans, comparable to research-grade equipment.

Robin's Viewer integrates deep-learning predictions into an EEG annotation tool to improve accuracy and allow user inspection of model outputs.

The paper introduces a method to estimate functional EEG sources using topographical templates, showing it performs as well as individual source localization methods but is more cost-effective and interpretable.

This study evaluates the use of resting state EEG for Parkinson's disease classification, finding that fewer electrodes and open-eye conditions improve accuracy.

This study compares RNNs to other neural networks in real-time decoding of finger movements from nonhuman primate neural signals, finding LSTMs outperform convolutional and transformer-based models.

This paper shows that deep learning models for detecting interictal epileptiform discharges in EEGs perform differently based on the type of EEG data used for training.

This paper addresses cross-subject sleep staging by applying contrastive learning based on inter-subject correlation, achieving state-of-the-art performance.

The paper shows longer estimation windows are better for EEG data analysis with overlap correction and provides an interactive app to visualize parameters.

The paper investigates whether switching tasks in an event-related design affects ERP amplitudes and decoding accuracies, finding no between-task correlations.

The paper applies single-trial-based temporal principal component analysis to extract event-related potentials from individual subject data, showing improved consistency with fewer trials.

HArtMuT improves M/EEG signal localization by modeling eye and muscle contributions, achieving comparable accuracy to brain sources.

This paper presents an ensemble algorithm that automatically classifies sleep stages using an optimized minimal number of EEG channels, achieving high accuracy with only 3 channels.

This paper describes a modular, adaptive, deep-learning-based Brain-VR Interface that integrates motor imagery BCI control into a gamified VR scenario to enhance long-term user motivation.

The paper introduces a large EEG dataset for visual object recognition and demonstrates its utility through multiple computational modeling approaches.

This paper develops a biologically-inspired neural network decoder for real-time brain-machine interfaces, achieving faster and more natural prosthetic finger movements than existing methods.