A novel EEG setup for online phase detection

Introduction: Local phase information of oscillatory activity has been shown to reflect different states of brain activity and is thought to play a crucial part in human brain information processing. In the visual system, phase properties of parieto-occipital alpha band oscillations were associated with perception and perception thresholds respectively. Along these lines, occipital TMS-induced phosphene-probability was related to the ongoing occipital alpha phase.

Purpose: Though phase information is thought to play an important role in signal processing in the human brain, there is few data addressing 'online' (real-time phase detection). Closing this gap, we evaluated a novel EEG setup utilizing a custom-made 10 channel EEG-recording system and low latency signal processing pipeline.

Material and Methods: In twelve healthy subjects (mean age 26.8 ± 2.6, 9 females) online phase detection was performed using individual spectral alpha peak adjusted, infinite impulse response bandpass filters and Hilbert transformation (electrode P3). Whenever phase angle of filtered signal was crossing 0 ° a marker was sent to a parallel recorded conventional EEG (250 trials, Brain Products, neighboured electrode P1). To contrast phase specific triggering another 250 trials with randomly distributed phase angles were detected in the same way.

Results: Angular distributions expressed by directedness (length of the mean vector) of electrode recordings at electrode P1 at the marker time point were significantly different from normal distribution (Rayleigh-Z highly significant) in all twelve subjects. In contrast random trials did not show any significant difference in directedness from normal distribution (Rayleigh-Z not significant in all subjects). Directedness was highly variable between subjects and correlated with individually filtered spectral alpha power in a linear regression analysis (p < 0.0001).

Summary: We evaluated precision of online-performance of a novel device to determine phase information within real-time by parallel EEG recordings with a conventional system. This approach allowed to show a good real-time determination of directed phases (directedness) at the respective marker points. Regression analysis revealed a correlation between respective spectral power and the magnitude of directedness. As next steps the novel 'online' device could be used to both present stimuli and trigger non-invasive brain stimulation in a phase-locked manner.

Fig. 1: Power spectra and angular distributions in two of twelve healthy volunteers. Raw power spectra, individual adjusted filter amplitude responses and filtered spectra are plotted on the left-hand side. Individual adjusted filter response thereby covers alpha amplitude peaks. On the right-hand side plots show angular distributions at marker time. Subject one shows small alpha amplitude and a low value of directedness. In contrast subject four shows high alpha amplitude and a high value of directedness.