Estimation of Interferences in Magnetoencephalography (MEG) Brain Data Using Intelligent Methods for BCI-based Neurorehabilitation Applications

Abstract

Brain-Computer Interface (BCI) neurorehabilitation offers the potential to improve recovery and quality of life for stroke survivors. It aims to restore lost physical and mental abilities through motor and cognitive therapies. Magnetoencephalography (MEG) signals are a major advancement in BCI technology as they provide accurate and consistent assessments of brain activity for control and interaction applications. MEG is indispensable for recording the magnetic fields produced in the brain during motor imagery tasks due to its capability to evaluate cerebral activity with remarkable temporal resolution. However, one of the major challenges associated with MEG recording is the loss of signal quality due to physiological artifacts and ambient noise. Additionally, the head movement of the individual during the recording process can result in the introduction of artifacts into the recorded data, which can distort the spatial mapping of brain activity. This, in turn, can jeopardize the reliability and accuracy of the results obtained. This study aims to identify the most effective technique for removing artifacts from MEG signals by conducting a comparative performance analysis of prominent denoising algorithms, such as Infomax, FastICA, SOBI, and SWT. The findings conclude that Infomax is the most effective algorithm for removing physiological artifacts from a signal while maintaining the integrity and essential features of the original data. FastICA was found to be the second most effective algorithm. Infomax outperformed FastICA in Power Spectral Density (PSD) and Percentage Root mean square error Difference (PRD) measurements.