A systematic review on artifact removal and classification techniques for enhanced MEG-based BCI systems

Neurological disease victims may be completely paralyzed and unable to move, but they may still be able to think. Their brain activity is the only means by which they can interact with their environment. Brain-Computer Interface (BCI) research attempts to create tools that support subjects with disabilities. Furthermore, BCI research has expanded rapidly over the past few decades as a result of the interest in creating a new kind of human-to-machine communication. As magnetoencephalography (MEG) has superior spatial and temporal resolution than other approaches, it is being utilized to measure brain activity non-invasively. The recorded signal includes signals related to brain activity as well as noise and artifacts from numerous sources. MEG can have a low signal-to-noise ratio because the magnetic fields generated by cortical activity are small compared to other artifacts and noise. By using the right techniques for noise and artifact detection and removal, the signal-to-noise ratio can be increased. This article analyses various methods for removing artifacts as well as classification strategies. Additionally, this offers a study of the influence of Deep Learning models on the BCI system. Furthermore, the various challenges in collecting and analyzing MEG signals as well as possible study fields in MEG-based BCI are examined.