The automatic detection of artifacts in Magnetoencephalographic signals based on fast ICA

As a non-invasive technique applied for the functional mapping of human brain, Magnetoencephalography (MEG) can acquire the neural activity with high temporal resolution and moderate spatial resolution. However, when reading a MEG record, for research or clinical reference, the investigator face the signals from non-cerebral sources like eye movements, heart beat and muscle activity always appear mixed with brain signals. In this article, we proposed a procedure including the independent component analysis(ICA) followed by an automatic independent component(IC) detection module mainly based on the analysis of statistical and spectral characteristics of each IC to remove the artifacts from MEG signals. The whole process was tested with both simulated data and real MEG signal, the results showed that the proposed technique was able to differentiate artfactual ICs successfully.