|
|
High-density resting-state EEG open-access data: Current status, challenges, and future perspectives
GUO Yatong, HU Jingyi, LEI Xu
2025, 33 (9):
1575-1591.
doi: 10.3724/SP.J.1042.2025.1575
The field of cognitive neuroscience has made significant strides through the use of high-density resting-state electroencephalography (EEG), a non-invasive technique that provides a unique window into the brain's intrinsic activity. This study systematically examines the current landscape, challenges, and future prospects of open-access high-density resting-state EEG data, highlighting its critical role in advancing our understanding of neural mechanisms underlying cognitive functions and psychiatric disorders.
Resting-state EEG (rsEEG) has emerged as a powerful tool due to its simplicity, cost-effectiveness, and high temporal resolution, allowing researchers to capture the brain's spontaneous neural oscillations. These oscillations, analyzed through specific frequency bands, have been linked to various cognitive processes and behaviors. Notably, rsEEG has shown significant potential in identifying biomarkers for mental illnesses, contributing to both fundamental research and clinical applications. However, existing datasets, predominantly sourced from Western, educated, industrialized, rich, and democratic (WEIRD) populations, exhibit limitations in geographic diversity and population coverage. The majority of shared datasets originate from Europe and North America, with a notable scarcity of contributions from Africa, highlighting the need for more inclusive and diverse data collection to enhance the generalizability of findings.
This study systematically evaluates 30 publicly available high-density (≥60 electrodes) rsEEG datasets, revealing critical gaps in geographic diversity, longitudinal design, and multimodal integration. Notably, 73% of these datasets originate from Europe and North America, while Africa remains underrepresented, underscoring the urgent need for inclusive, globally representative data to address the WEIRD sample bias. Our analysis identifies key limitations in existing databases, such as the predominance of cross-sectional studies, which hinder investigations into neurodevelopmental trajectories and aging processes.
The application of rsEEG spans multiple domains, including the study of sleep deprivation effects, neurodevelopment, and the identification of biomarkers for neuropsychiatric disorders. In sleep research, rsEEG identifies predictors of outcomes after sleep deprivation. It also aids in building lifespan databases for neurodevelopment insights. Clinically, rsEEG detects biomarkers for Alzheimer's, autism, depression, epilepsy, and insomnia. Large datasets have laid the foundation for exploring disease-specific neural oscillations, underscoring the versatility of rsEEG in both clinical and research contexts.
A major innovation of this study lies in its detailed examination of emerging analytical methodologies that leverage high-density rsEEG data. We highlight the shift from traditional spectral and connectivity analyses to advanced techniques like aperiodic power spectrum analysis, which distinguishes periodic oscillations from nonperiodic neural activity, offering new insights into excitatory-inhibitory balance in neuropsychiatric disorders. Furthermore, we catalog cutting-edge open-source toolkits that standardize preprocessing and feature extraction, enabling large-scale, reproducible research. Our findings demonstrate that databases accompanied by dedicated description papers achieve significantly higher citation rates compared to those without, emphasizing the importance of scholarly documentation in promoting data reuse.
The integration of artificial intelligence (AI) with rsEEG represents another groundbreaking contribution. We review how deep learning models, such as DeprNet and HybridEEGNet, achieve unprecedented accuracy in diagnosing depression and Parkinson’s disease by automating feature extraction from raw EEG signals. Additionally, we introduce pioneering EEG foundation models trained on thousands of hours of data, which outperform traditional methods in tasks like epilepsy detection and sleep stage classification. These models address the critical challenge of limited training data through synthetic EEG generation techniques. Our discussion of AI extends to its role in democratizing EEG analysis, with tools like DISCOVER-EEG reducing reliance on manual preprocessing and subjective expert judgment.
Looking ahead, we propose a roadmap for advancing rsEEG research through FAIR (Findable, Accessible, Interoperable, Reusable) data-sharing practices and the adoption of EEG-BIDS standards. We advocate for multisite collaborations to build diverse, longitudinal cohorts spanning the lifespan, particularly targeting underrepresented populations and neurological conditions. The study also underscores the potential of wearable dry-electrode systems and edge-computing frameworks to enable real-time, large-scale rsEEG monitoring outside clinical settings. By addressing current limitations in data diversity, analytical robustness, and translational applications, this work lays the foundation for rsEEG to drive precision medicine and global brain health initiatives.
In conclusion, our study not only synthesizes the state-of-the-art in high-density rsEEG but also pioneers actionable strategies to harness its full potential. The innovations highlighted—from AI-driven diagnostics to equitable data governance—position rsEEG as an indispensable tool for unraveling the complexities of brain function and dysfunction in the coming decade.
References |
Related Articles |
Metrics
|
