Abstract:OBJECTIVE: Brain-computer interfaces (BCI) enable direct communication with a computer, using neural activity as the control signal. This neural signal is generally chosen from a variety of well-studied electroencephalogram (EEG) signals. For a given BCI paradigm, feature extractors and classifiers are tailored to the distinct characteristics of its expected EEG control signal, limiting its application to that specific signal. Convolutional neural networks (CNNs), which have been used in computer vision and speech recognition to perform automatic feature extraction and classification, have successfully been applied to EEG-based BCIs; however, they have mainly been applied to single BCI paradigms and thus it remains unclear how these architectures generalize to other paradigms. Here, we ask if we can design a single CNN architecture to accurately classify EEG signals from different BCI paradigms, while simultaneously being as compact as possible.APPROACH: In this work we introduce EEGNet, a compact convolutional neural network for EEG-based BCIs. We introduce the use of depthwise and separable convolutions to construct an EEG-specific model which encapsulates well-known EEG feature extraction concepts for BCI. We compare EEGNet, both for within-subject and cross-subject classification, to current state-of-the-art approaches across four BCI paradigms: P300 visual-evoked potentials, error-related negativity responses (ERN), movement-related cortical potentials (MRCP), and sensory motor rhythms (SMR).MAIN RESULTS: We show that EEGNet generalizes across paradigms better than, and achieves comparably high performance to, the reference algorithms when only limited training data is available across all tested paradigms. In addition, we demonstrate three different approaches to visualize the contents of a trained EEGNet model to enable interpretation of the learned features.SIGNIFICANCE: Our results suggest that EEGNet is robust enough to learn a wide variety of interpretable features over a range of BCI tasks. Our models can be found at: https://github.com/vlawhern/arl-eegmodels.

ExBrainable: An Open-Source GUI for CNN-based EEG Decoding and Model Interpretation

Portable Brain-Computer Interface Based on Novel Convolutional Neural Network.

EEGNet: a compact convolutional neural network for EEG-based brain–computer interfaces

Decoding visual brain representations from electroencephalography through Knowledge Distillation and latent diffusion models

Brain Dialogue Interface (BDI): A User-Friendly fMRI Model for Interactive Brain Decoding

EEGNet: A Compact Convolutional Network for EEG-based Brain-Computer Interfaces

Visual Explanations of Deep Convolutional Neural Network for EEG Brain Fingerprint

EEG2TEXT: Open Vocabulary EEG-to-Text Decoding with EEG Pre-Training and Multi-View Transformer

Toward reliable signals decoding for electroencephalogram: A benchmark study to EEGNeX

GCNs-Net: A Graph Convolutional Neural Network Approach for Decoding Time-resolved EEG Motor Imagery Signals

Ieegview: an Open-Source Multifunction GUI-based Matlab Toolbox for Localization and Visualization of Human Intracranial Electrodes.

Enhancing EEG Motor Imagery Decoding Performance Via Deep Temporal-domain Information Extraction

Neurodm: Decoding and Visualizing Human Brain Activity with Eeg-Guided Diffusion Model

Visual Decoding and Reconstruction via EEG Embeddings with Guided Diffusion

EEG Conformer: Convolutional Transformer for EEG Decoding and Visualization

Decoding human brain activity with deep learning

Efficient Brain Decoding Based on Adaptive EEG Channel Selection and Transformation

Neuro-3D: Towards 3D Visual Decoding from EEG Signals

Portable deep-learning decoder for motor imaginary EEG signals based on a novel compact convolutional neural network incorporating spatial-attention mechanism

BrainNet Viewer: A Network Visualization Tool for Human Brain Connectomics