Signal source localization from spatio-temporal biomagnetic data by signal subspace method

New methods are proposed to localize multiple current dipoles from spatio-temporal biomagnetic data obtained with SQUID magnetometers. The proposed methods are based on subspace fitting (SF) and weighted subspace fitting (WSF) frameworks that have been developed originally in the field of array signal processing. When time series from multiple dipoles are strongly correlated or multiple dipoles are closely spaced, these methods outperform the multiple signal classification (MUSIC) proposed by Mosher et al. [1]. Results are presented using these methods for simulated data and experimental magnetocardiogram data. The WSF criterion produced significantly better results compared with the MUSIC and the SF criteria.