Phase transfer entropy: A novel phase-based measure for directed connectivity in networks coupled by oscillatory interactions
Phase synchronization (PS) analysis has been demonstrated to be a useful method to infer functional connectivity with multichannel neural signals, e.g., electroencephalography (EEG). Methodological problems on quantifying functional connectivity with PS analysis have been investigated extensively, but some of them have not been fully solved yet. For example, how long a segment of EEG signal should be used in estimating PS index? Which methods are more suitable to infer the significant level of estimated PS index? To address these questions, this paper performs an intensive computation study on PS analysis based on surrogate tests with 1) artificial surrogate data generated by shuffling the rank order, the phase spectra, or the instantaneous frequency of original EEG signals, and 2) intersubject EEG pairs under the assumption that the EEG signals of different subjects are independent. Results show that 1) the phase-shuffled surrogate method is workable for significance test of estimated PS index and yields results similar to those by intersubject EEG surrogate test; 2) generally, a duration of EEG waves covering about 3 16 cycles is suitable for PS analysis; and 3) the PS index based on mean phase coherence is more suitable for PS analysis of EEG signals recorded at relatively low sampling rate.