MEG recordings of DC fields using the signal space separation method (SSS).

Abstract

Stationary SQUID sensors record time-varying magnetic fields only. Any DC sources, such as magnetic impurities on the scalp or physiological DC currents, are invisible in conventional MEG with stationary sources and sensors. However, movement of the subject relative to the measurement device transforms the DC fields into time-varying MEG signals, which are either signals of interest from biomagnetic sources, or movement artifacts when caused by magnetic residue on the head. These signals can be demodulated to DC by tracking the head movement and by using this recorded information to decompose the signals into a device-independent source model. To do this we have used the signal space separation method (SSS) along with a continuous head position monitoring system. From time variations of the recorded signal, a linear equation is obtained relating the averaged MEG signal variation, the DC-source in the head, and the varying external interference. In this way an unbiased estimate is obtained for the DC source as it is automatically separated from external interference. The method was tested by feeding DC current in an artificial current dipole on a phantom head and by continuously moving and rotating this phantom randomly with a motion amplitude of several centimeters. After the SSS based movement demodulation and reconstruction of the signal from inside of the helmet, the location of the DC current dipole in the phantom could be determined with an accuracy of 2 mm. It is concluded that the method enables localization of DC sources with MEG using voluntary head movements.