Discriminating Multiple Source Components of Magnetoencephalogram by Time-frequency Analysis
M. Sano, Kohei Tanaka, Y. Uchikawa, S. Sakurai, Toshiro Watanabe, B.S. Kim, K. Kobayashi · Journal of the Magnetics Society of Japan · 2009
We describe a new method for discriminating multiple source overlapping in time by using time-frequency information. We measured magnetic fields evoked by somatosensory, auditory and mixed stimulation. We reduced noise by independent component analysis and detected the characteristic frequencies of auditory and somatosensory evoked responses by time-frequency analysis. We then discriminated auditory and somatosensory components from the compound MEF response waveforms by applying narrow-band-pass filters having band-widths of 10 Hz with pass-bands centered at those characteristic frequencies for auditory and somatosensory respectively. The time-frequency maps in discriminated results agree with the map obtained from independent stimulation. These results indicate that proposed method is useful for discriminating multiple sources overlapping in time.