Reducing the Effect of Spurious Phase Variations in Neural Oscillatory Signals
Published on Oct 8, 2018in Frontiers in Computational Neuroscience2.32
· DOI :10.3389/fncom.2018.00082
The phase-reset model of oscillatory EEG activity has received a lot of attention in the last decades for decoding different cognitive processes. Based on this model, the ERPs are assumed to be generated as a result of phase reorganization in ongoing EEG. In addition, the study of oscillatory EEG signals can be used to overcome limitations regarding the study of segmented EEG data, i.e., ERPs. Measuring the level of instantaneous phase (IP) synchronization has been used in numerous studies of ERPs as well as oscillatory activity for a better understanding of the underlying neural activities. However, the reliability of results can be challenged as a result of noise artefact in IP. Phase distortion due to environmental noise artifacts as well as different pre-processing steps on signals can lead to generation of artificial phase jumps. One of such effects presented recently is the effect of low envelope on the IP of signal. It has been shown that as the instantaneous envelope of the analytic signal approaches zero, the variations in the phase increase, effectively leading to abrupt transitions in the phase. These abrupt transitions can distort the phase synchronization results as they are not related to any neurophysiological effect. These transitions are called spurious phase variation. In this study, we present a model to remove generated artificial phase variations due to the effect of low envelope. The proposed method is based on a simplified form of a Kalman smoother, that is able to model the IP behavior in narrow-bandpassed oscillatory signals. The method is not only evaluated on synthetic data but also in experimental EEG measurements recorded using a listening dichotic paradigm designed to assess auditory selective attention between an attended and unattended conditions.