Real-time EEG Forecasting and Phase-locked, ‘Closed-loop’ tACS

Mansouri, Farrokh 1 ; Downar, Jonathan 3 ; Zariffa, José 1, 2

1. Institute of Biomaterial and Biomedical Engineering, University of Toronto, Toronto, Ontario, Canada; 2. Toronto Rehab Institute, University Health Network, Toronto, Ontario, Canada; 3. Department of Psychiatry, University of Toronto, Toronto, Ontario, Canada


A growing body of research is showing that noninvasive electrical brain stimulation phase-locked to the underlying brain rhythms can effectively modulate neural activity, with a wide range of applications in treating brain disorders. Transcranial alternating current stimulation(tACS) can stimulate the brain in-phase to its natural oscillations as recorded by electroencephalography(EEG), but matching these oscillations in real-time is challenging, due to the complex and time-varying nature of the EEG. This work addresses this challenge by investigating an algorithm to control 'closed-loop' tACS stimulation in real-time.


Our algorithm extracts phase and frequency from a segment of EEG, then forecasts the signal to control the stimulation in real-time. A careful tuning of the EEG segment length and prediction horizon is required and has been investigated here for different EEG frequency bands. The algorithm was tested on EEG from 5 healthy volunteers. The performance of this algorithm was quantified through phase-locking to various EEG frequency bands.


The developed algorithm performs faster and provides better phase-locked stimulation compared to the current state-of-the-art. Phase-locking performance was consistent across individuals and recording locations. With current parameters, the algorithm performs best when tracking oscillations in the alpha band (8-14Hz), with a phase-locking value of 0.77+0.08. Performance was further found to be maximized when the frequency band of interest has a dominant frequency that is stable over time. This algorithm will be used in upcoming studies exploring the effects of tACS in psychiatric conditions.


In this work, a new frequency specific forecasting algorithm for EEG signal was proposed and tested. Based on the analysis done on synthetic and recorded EEG, the algorithm is sufficiently fast and can provide good phase locking, making it suitable for closed-loop, phase-locked brain tACS.