Non-stationarity Removal Techniques in MEG Data: A Review

Brain Computer Interface (BCI) enables communication solely through mental activity. For patients who have lost complete voluntary muscle control, it serves as a potential communication and rehabilitation channel. In a number of recent investigations, BCl based on electroencephalography (EEG) has demonstrated increased reaction in nonresponsive people to interact with others. Despite the method's effectiveness, patient interaction takes way too long. However, magnetoencephalography (MEG) may shorten the training period, improving BCI reliability in the process. One of the major technological difficulties confronting MEG data collection and interpretation is that the strength of neuromagnetic fields recorded externally is considerably lower than interfering signals. There hasn't been much substantial progress on an effective MEG-BCI system because there is no large enough MEG dataset. Despite their huge potential, MEG-based BCI systems still need a lot of work in terms of signal processing algorithms that are both reliable and efficient. Unexpected head movements and changing orientation between sessions may be the cause of non-stationarity in the MEG data that was captured, changing the most effective channel selection between sessions and participants. Not only the head movement, but also non-stationarity in data can be caused by a variety of factors like user weariness, mood changes, or external noise interfering with the MEG system. This paper discusses the many types of data acquisition artifacts and review the techniques to minimize the artifacts to increase the signal-to-noise ratio.