Multiple dipole sources identification from an EEG topography using information criteria

The electric activity in the human cerebral cortex can be recorded with surface EEG electrodes applied to the scalp. The source of recorded EEG signals can be approximated to one or more equivalent current dipoles within the brain. It is an important problem that how to determine the optimal dipole number. In this paper, we propose a new method combining the Powell algorithm and the information criterion method for determining the optimal dipole number. With the common model, it is shown how to calculate the potential error by the Powell algorithm with the cost function, and how to use this potential error to choose the optimal dipole number by the information criterion method. The new method has the advantages of identification accuracy of dipole number and EEG data number, because in this method: (1) only an EEG topography is used in the computation, (2) the information criterion method can get the high accuracy. In order to prove our method to be efficient, precise and robust to the noise, the 10% white noise inserted to test this method. Results are presented here to show our method is an efficient approach for determining the dipole number.