Unsupervised identification of spontaneous magnetoencephalographic alpha activity by Independent Component Analysis

摘要

Information processing in the brain involves oscillatory electrical signals which can be measured using multichannel electro- or magnetoencephalography (MEG). Best studied are occipital alpha- and pericentral mu-rhythms with frequencies at 10Hz and its higher harmonics. Here, we analysed continuous MEG recordings from a visual memory task using Independent Component Analysis (ICA), which separates statistically independent sources in multichannel recordings without a priori knowledge. ICA isolated several oscillatory brain sources. A subsequent localisation using the dipole in a conducting sphere model gave the location of the sources on the cortex with a high goodness of fit. One of the sources could be identified as pericentral mu-oscillations due to a suppression and a rebound of the oscillations triggered by motor activity.