Deep learning-based electroencephalic decoding of the phase-lagged transcranial alternating current stimulation

Area of Science:

• Neuroscience
• Cognitive Science
• Biomedical Engineering

Background:

• Non-invasive brain stimulation techniques like cross-frequency coupled alternating current stimulation (CFC-tACS) aim to enhance cognitive functions.
• Understanding the brain's response to different stimulation parameters is crucial for optimizing these techniques.
• Electroencephalography (EEG) is a widely used tool for measuring brain activity.

Purpose of the Study:

• To investigate if task-based EEG signals can decode different phase-lag types of online CFC-tACS.
• To explore the potential of using EEG to monitor and decode brain states during neuromodulation.
• To assess the feasibility of closed-loop brain-machine interfacing (BMI) systems incorporating neuromodulation.

Main Methods:

• EEG data were collected from 21 healthy participants performing a modified Sternberg task.
• CFC-tACS with 45° or 180° phase lags between the central executive network and default mode network was applied online.
• A modified EEGNet model was trained on EEG signals to decode the applied CFC-tACS conditions.

Main Results:

• The EEGNet model achieved an 81.73% decoding accuracy using parietal EEG signals.
• Feature maps indicated that the model primarily relied on EEG beta activity in the parietal region.
• This suggests that tACS phase-lag type influences top-down cognitive control, reflected in beta-band activity.

Conclusions:

• Task-based EEG signals are capable of decoding online CFC-tACS stimulation types.
• Task-related EEG spectral characteristics, particularly beta activity, can serve as indicators of neuromodulatory effects.
• This research supports the advancement of communicative strategies in closed-loop BMI-neuromodulation systems.