Predicting spiking activity from scalp EEG

Area of Science:

• Neuroscience
• Computational Neuroscience
• Biomedical Engineering

Background:

• The relationship between electroencephalography (EEG) and neural spiking activity is not fully understood.
• This knowledge gap hinders the inference of neural dynamics from non-invasive EEG and the development of brain-machine interfaces (BMIs).

Purpose of the Study:

• To estimate spiking activity in the visual cortex using non-invasive scalp EEG.
• To investigate the predictive power of EEG spectrotemporal features for neural spiking activity.

Main Methods:

• Recorded simultaneous scalp EEG and V1 multi-unit activity envelope (MUAe) in a macaque monkey.
• Analyzed the relationship between MUAe and EEG frequency bands, extracting phase, amplitude, and phase-amplitude coupling.
• Used linear regression to predict MUAe from EEG features.

Main Results:

• Spectrotemporal EEG features reliably predicted V1 MUAe, despite complex and frequency-dependent relationships.
• EEG phase, amplitude, and coupling each contributed to MUAe prediction accuracy.
• MUAe prediction was more accurate in superficial cortical layers, and stimulus frequency phase further improved predictions.

Conclusions:

• Comprehensive spectrotemporal features of non-invasive EEG contain significant information about underlying spiking activity.
• This finding highlights the richness of EEG signals and their complex relationship with neural dynamics.
• Utilizing detailed spectrotemporal EEG signatures can enhance the performance of BMI applications.