Robust learning from corrupted EEG with dynamic spatial filtering.

Building machine learning models using EEG recorded outside of the laboratory setting requires methods robust to noisy data and randomly missing channels. This need is particularly great when working with sparse EEG montages (1-6 channels), often encountered in consumer-grade or mobile EEG devices. Neither classical machine learning models nor deep neural networks trained end-to-end on EEG are typically designed or tested for robustness to corruption, and especially to randomly missing channels. While some studies have proposed strategies for using data with missing channels, these approaches are not practical when sparse montages are used and computing power is limited (e.g., wearables, cell phones). To tackle this problem, we propose dynamic spatial filtering (DSF), a multi-head attention module that can be plugged in before the first layer of a neural network to handle missing EEG channels by learning to focus on good channels and to ignore bad ones. We tested DSF on public EEG data encompassing ∼4000 recordings with simulated channel corruption and on a private dataset of ∼100 at-home recordings of mobile EEG with natural corruption. Our proposed approach achieves the same performance as baseline models when no noise is applied, but outperforms baselines by as much as 29.4% accuracy when significant channel corruption is present. Moreover, DSF outputs are interpretable, making it possible to monitor the effective channel importance in real-time. This approach has the potential to enable the analysis of EEG in challenging settings where channel corruption hampers the reading of brain signals.

Characterizing Population EEG Dynamics throughout Adulthood.

For decades, electroencephalography (EEG) has been a useful tool for investigating the neural mechanisms underlying human psychological processes. However, the amount of time needed to gather EEG data means that most laboratory studies use relatively small sample sizes. Using the Muse, a portable and wireless four-channel EEG headband, we obtained EEG recordings from 6029 subjects 18-88 years in age while they completed a category exemplar task followed by a meditation exercise. Here, we report age-related changes in EEG power at a fine chronological scale for δ, θ, α, and ß bands, as well as peak α frequency and α asymmetry measures for both frontal and temporoparietal sites. We found that EEG power changed as a function of age, and that the age-related changes depended on sex and frequency band. We found an overall age-related shift in band power from lower to higher frequencies, especially for females. We also found a gradual, year-by-year slowing of the peak α frequency with increasing age. Finally, our analysis of α asymmetry revealed greater relative right frontal activity. Our results replicate several previous age- and sex-related findings and show how some previously observed changes during childhood extend throughout the lifespan. Unlike previous age-related EEG studies that were limited by sample size and restricted age ranges, our work highlights the advantage of using large, representative samples to address questions about developmental brain changes. We discuss our findings in terms of their relevance to attentional processes and brain-based models of emotional well-being and aging.