Cortical areas classification via AR modeling and 3-D spectral estimation

Magnetoencephalogram (MEG) signals are processed via autoregressive (AR) modeling and 3-D spectral estimation. The Ulrich-Clayton method along with the technique of signal averaging satisfactorily describes the data. The order of the AR filter depends on the distance of the recording point on the scalp from the acoustic center. The variations of power distribution of MEG signals due to the application of stimuli are examined via 3-D spectral estimation. Simple implementation and data compression properties make AR modeling suitable for clinical application. Both methods can be used to locate regions of the brain which do not function properly.<>