Detecting single-trial P300 from EEG is difficult when only a few labeled trials are available. When attempting to boost a small target set with a large source dataset through transfer learning, cross-dataset shift arises. To address this challenge, we study transfer between two public visual-oddball ERP datasets using five shared electrodes (Fz, Pz, P3, P4, Oz) under a strict small-sample regime (target: 10 trials/subject; source: 80 trials/subject). We introduce Adaptive Split Maximum Mean Discrepancy Training (AS-MMD), which combines (i) a target-weighted loss with warm-up tied to the square root of the source/target size ratio, (ii) Split Batch Normalization (Split-BN) with shared affine parameters and per-domain running statistics, and (iii) a parameter-free logit-level Radial Basis Function kernel Maximum Mean Discrepancy (RBF-MMD) term using the median-bandwidth heuristic. Implemented on an EEG Conformer, AS-MMD is backbone-agnostic and leaves the inference-time model unchanged. Across both transfer directions, it outperforms target-only and pooled training (Active Visual Oddball: accuracy/AUC 0.66/0.74; ERP CORE P3: 0.61/0.65), with gains over pooling significant under corrected paired t-tests. Ablations attribute improvements to all three components.

ICLabel is an important plug-in function in EEGLAB, the most widely used software for EEG data processing. A powerful approach to automated processing of EEG data involves decomposing the data by Independent Component Analysis (ICA) and then classifying the resulting independent components (ICs) using ICLabel. While EEGLAB pipelines support high-performance computing (HPC) platforms running the open-source Octave interpreter, the ICLabel plug-in is incompatible with Octave because of its specialized neural network architecture. To enhance cross-platform compatibility, we developed a Python version of ICLabel that uses standard EEGLAB data structures. We compared ICLabel MATLAB and Python implementations to data from 14 subjects. ICLabel returns the likelihood of classification in 7 classes of components for each ICA component. The returned IC classifications were virtually identical between Python and MATLAB, with differences in classification percentage below 0.001%.

The rise of open-source neuroimaging data sharing has created new opportunities for advancing machine learning in neuroscience. However, this trend also highlights the need for a framework capable of leveraging these large-scale, publicly available datasets for advanced machine learning algorithms, particularly deep learning techniques. In this paper, we introduce EEG-SSL, a framework designed to work with large heterogenous EEG neuroimaging datasets shared in the standardized Brain Imaging Data Structure (BIDS) format using Self-Supervised Learning (SSL). SSL is a powerful deep-learning approach that benefits from large-scale datasets, even when unlabeled or minimally labeled. We show how information can be extracted from standardized EEG BIDS-formated data and discuss different considerations while processing at scale in our framework. We specify the software components that are needed to apply SSL to EEG and provide several standard implementations for each of the components. We also describe how each component can be extended with custom implementation while leveraging the supporting functions of the framework. We applied the framework to a large-scale BIDS-compliant dataset (ds004186) comprising resting-state EEG data from over 2,000 subjects, showcasing how SSL can be structured and deployed for potential big data in EEG.

EEG continues to find a multitude of uses in both neuroscience research and medical practice, and independent component analysis (ICA) continues to be an important tool for analyzing EEG. A multitude of ICA algorithms for EEG decomposition exist, and in the past, their relative effectiveness has been studied. AMICA is considered the benchmark against which to compare the performance of other ICA algorithms for EEG decomposition. AMICA exposes many parameters to the user to allow for precise control of the decomposition. However, several of the parameters currently tend to be set according to “rules of thumb” shared in the EEG community. Here, 70-channel AMICA decompositions are run on data from a collection of participants while varying certain key parameters. The running time and quality of decompositions are analyzed based on two metrics, Pairwise Mutual Information (PMI) and Mutual Information Reduction (MIR), and derived recommendations for selecting parameter values are presented.

Deep learning models are capable of extracting task-relevant information in complex brain dynamics from large corpora of raw EEG data. Given the small size of EEG datasets and the heterogeneity of each dataset’s channel montage, aggregating EEG datasets for large-scale training of deep learning models often requires a way to harmonize different channel locations effectively. Previous methods have focused on extracting features from raw EEG and projecting them onto a common space. However, these approaches underexploit the potential richness of EEG raw data. Here, we proposed a method to train deep learning models on raw EEG data with different channel montages using spatial attention on electrode coordinates. We test this approach on a gender classification task. We first show that increasing channel montage density increases performance. We then show that spatial attention increases model performance for different channel montage densities. Then, we show that in nonuniform montage settings, deep learning models trained on combined data with different channel montages performs significantly better than deep learning models trained on individual datasets with fixed montages. This work opens up the potential for future application of deep learning models on datasets collected by different labs across different recording settings.

Independent component analysis (ICA), is a blind source separation method that is becoming increasingly used to separate brain and non-brain related activities in electroencephalographic (EEG) and other electrophysiological recordings. It can be used to extract effective brain source activities and estimate their cortical source areas, and is commonly used in machine learning applications to classify EEG artifacts. Previously, we compared results of decomposing 1371-channel scalp EEG datasets using 22 ICA and other blind source separation (BSS) algorithms. We are now making this framework available to the scientific community and, in the process of its release are testing a recent ICA algorithm (Picard) not included in the previous assay. Our test framework uses three main metrics to assess BSS performance: Pairwise Mutual Information (PMI) between scalp channel pairs; PMI remaining between component pairs after decomposition; and, the complete (not pairwise) Mutual Information Reduction (MIR) produced by each algorithm. We also measure the “dipolarity” of the scalp projection maps for the decomposed component, defined by the number of components whose scalp projection maps nearly match the projection of a single equivalent dipole located in the volume of a template boundary element method (BEM) electrical forward problem head model. Within this framework, Picard performed similarly to Infomax ICA. This is not surprising since Picard is a type of Infomax algorithm that uses the LBFGS method for faster convergence, in contrast to Infomax and Extended Infomax (runica) which use gradient descent. Our results show that Picard performs similarly to Infomax and, likewise, better than other BSS algorithms, excepting the more computationally complex AMICA. Further research might determine if partial Picard decomposition, followed by AMICA, might produce unequaled performance without a large time penalty. We have released the source code of our framework and the test data through GitHub to encourage further comparisons of ICA/BSS algorithm performance applied to electrophysiological data (https://github.con/sccn/testica).

EEG power spectral density (PSD), the individual alpha frequency (IAF) and the frontal alpha asymmetry (FAA) are all EEG spectral measures that have been widely used to evaluate cognitive and attentional processes in experimental and clinical settings, and that can be used for real-world applications (e.g., remote EEG monitoring, brain-computer interfaces, neurofeedback, neuromodulation, etc.). Potential applications remain limited by the high cost, low mobility, and long preparation times associated with high-density EEG recording systems. Low-density wearable systems address these issues and can increase access to larger and diversified samples. The present study tested whether a low-cost, 4-channel wearable EEG system (the MUSE) could be used to quickly measure continuous EEG data, yielding similar frequency components compared to a research-grade EEG system (the 64-channel BIOSEMI Active Two). MUSE data can be live-streamed using the Lab Stream Layer (LSL), and can therefore be implemented into real-world EEG monitoring, brain-computer interfaces (BCI), or neurofeedback applications. We compare the spectral measures from MUSE EEG data referenced to mastoids to those from BIOSEMI EEG data with two different references for validation (mastoids and average reference). A minimal amount of data was deliberately collected to test the feasibility for real-world applications (EEG setup and data collection being completed in under 5 min). We show that the MUSE can be used to examine power spectral density (PSD) in all frequency bands, the individual alpha frequency (IAF), and frontal alpha asymmetry (FAA). Furthermore, we observed satisfying internal consistency reliability in alpha power and asymmetry measures recorded with the MUSE. However, estimating asymmetry on the IAF did not yield significant advantages relative to the traditional method (average over the 8-13 Hz range). These findings should advance human neurophysiological monitoring using easily accessible wearable neurotechnologies in large samples and increase the feasibility of their implementation in real-world settings.

Wearable EEG headsets have transformed the landscape of EEG research. It is no longer necessary to use expensive equipment and over 30 minutes of preparation time to collect EEG data. Instead, participants may do it themselves in a few minutes from the comfort of their home. Confronted with processing thousands of such recordings, manual data cleaning has become a bottleneck, so we tested automated methods for cleaning data. To validate these methods, we asked three trained EEG human raters to clean 100 files of 12 minutes each, and use these manual rejections to assess the performance of automated cleaning methods for both channel rejections and continuous data rejections. We showed that rejecting channels based on abnormal spectrum yielded the best results. We also showed that the Artifact Subspace Reconstruction rejection method was the best method to reject continuous portions of data. Inter-rater consistency is the gold standard to assess the quality of automated data rejection methods, and we showed that our best rejection methods were not significantly different and might even outperform human raters. We provide a simple recipe and plugin for the popular EEGLAB software for automated data cleaning of Muse data. We hope this new tool will allow more widespread use of wearable EEG in clinical and research settings where large quantities of wearable EEG data need to be processed.

Convolutional Neural Networks (CNNs) have achieved impressive performance on many computer vision-related tasks, such as object detection, image recognition, image retrieval, etc. These achievements benefit from the CNNs’ outstanding capability to learn discriminative features with deep layers of neuron structures and iterative training processes. This has inspired the EEG research community to adopt CNN in performing EEG classification tasks. However, CNNs learned features are not immediately interpretable, causing a lack of understanding of the CNNs’ internal working mechanism. To improve CNN interpretability, CNN visualization methods are applied to translate the internal features into visually perceptible patterns for qualitative analysis of CNN layers. Many CNN visualization methods have been proposed in the Computer Vision literature to interpret the CNN network structure, operation, and semantic concept, yet applications to EEG data analysis have been limited. In this work we use 3 different methods to extract EEG-relevant features from a CNN trained on raw EEG data: optimal samples for each classification category, activation maximization, and reverse convolution. We applied these methods to a high-performing Deep Learning model with state-of-the-art performance for an EEG sex classification task, and show that the model exploits differences between classes in the theta frequency band. We show that the visualization of a CNN model can reveal interesting EEG biomarkers. Using these tools, EEG researchers using Deep Learning can better identify the learned EEG features, possibly identifying new class-relevant biomarkers.

Phase-Amplitude Coupling (PAC) in electrophysiological signals refers to the transient interplay of activities in different frequency ranges, wherein phase in a low-frequency band and amplitude in a high-frequency band are in some way dependent. PAC phenomena have received increasing interest in neuroscience given the growing evidence of their apparent role in both normal and pathological brain processes. This interest has resulted in publication of a wave of methods for PAC estimation, each with its own advantages and drawbacks relative to others. Motivated by the widespread study of this phenomenon, most academic open source software environments for analyzing electrophysiological signals (e.g., Fieldtrip, Brainstorm, MNE) have implemented at least one method of PAC estimation. Here we describe an EEGLAB plug-in release, PACTools for MATLAB (The Mathworks, Inc.), that computes PAC in either continuous or event-related data using any of five methods for PAC estimation: Mean Vector Length Modulation Index, Kullback- Leibler Modulation Index, Phase-Locking Value, General Linear Model Modulation Index and Mutual Information PAC (MIPAC). PACTools uses parallelized code for efficient performance and offers built-in direct access to online high-performance computing resources made freely available for nonprofit research by the Neurosciences Gateway (nsgportal.org). PACTools features intuitive graphic user interfaces and equivalent command line calls that make its use straightforward and seamless in the EEGLAB environment. We discuss toolbox implementation, architecture, dependencies, and the implemented methods of PAC computation and visualization.

Recent research has shown that the speed of image processing achieved by the human visual system is incompatible with conventional neural network approaches that use standard coding schemes based on firing rate. An alternative is to use networks of asynchronously firing spiking neurons and use the order of firing across a population of neurons as a code. In this paper we summarize results that demonstrate a number of advantages of such coding schemes: (1) they allow very efficient transmission of information, (2) they are intrinsically invariant to variations in stimulus intensity and contrast, (3) they can be used in very large scale processing architectures to solve difficult problems including categorization of objects in natural scenes, and (4) they are particularly suited for implementation in low-cost multi-processor hardware.