Improving the performance of the signal space separation method by comprehensive spatial sampling.

Biomagnetic instruments usually employ sensors with approximately radial normal vectors arranged on a near-spherical surface. The multipole expansion employed in the recently introduced signal space separation method (SSS) reveals limitations in this traditional sensor array design. Specifically, we show that the excessive symmetry of the sensor array impedes separation of multipole components arising from inside and outside of the array. This motivates consideration of novel instrument designs that would sample the field in a more comprehensive way. We evaluated several simulated sensor arrays that employ vector sensors in one or two layers, giving information on multiple field components and the radial dependence of the field. Our results indicate that this kind of sensor array design could significantly improve SSS performance, leading to enhanced shielding against external interference and reduced noise after signal reconstruction. The best two-layer array evaluated here attains a shielding factor of nearly 1000 or 60 dB with about 400 sensors. Due to limited spatial coverage, a traditional reference array geometry does not give the same level of improvement. In addition to improved software shielding, enhanced detection of different multipole components increases the information obtained about the magnetic field, which has fundamental importance.