Single-trial Quantification of Imagery Beta-band Mu Rhythm in Finger Lifting Task Using Independent Component Analysis (ICA)

摘要

We propose an ICA-based method to quantify the single-trial response of the imagery beta-band event-related synchronization (ERS) in right hand finger lifting task. One right-handed female subject participated to perform imagery index finger lifting task. The magnetic fields were recorded by a whole-head magnetoencephalography (Neuromag-306). ICA was applied to separate each rhythmic epoch into temporally independent components by means of an un-mixing matrix. Each column of the inverse of un-mixing matrix represents a spatial map describing the relative projection weights of the corresponding temporal components at each of the MEG sensors. According to the temporal and spatial characteristics, the scalp map and temporally independent components can be categorized into two groups, one is task-related and the other is task-unrelated. The task-related components of each epoch were selected to reconstruct the data that was then band-passed filtered (around 16-20Hz) followed by the Hilbert transform to produce a beta-band envelope. The amplitude rebound of each single-trial beta band envelope was computed. The single-trial amplitude rebounds of beta band obtained from 100 epochs were 35.94±15.24fT/cm. Compared to the 100-trial averaged results obtained from the Amplitude Modulation (AM) method (Clochon et al., 1996), which was 25.18fT/cm, the mean value of the amplitude rebound in ICA-based method was higher since the ICA-based method removed the external and internal noises and thus improved the contrast of amplitude rebound.