Synchronization Tomography: Model Calculations of Spatial Resolution and Noise Tolerance; Optimization of Source Localization

Phase synchronization plays an important role both under physiological and pathological conditions. With standard averaging techniques of MEG data, it is difficult to reliably detect cortico-cortical and cortico-muscular phase synchronization processes that are not time-locked to an external stimulus. For this reason, novel synchronization analysis techniques were developed and directly applied to MEG signals [1]. However, due to the lack of an inverse modeling (i.e., source localization), the spatial resolution of this approach is limited. To detect and localize cerebral phase synchronization, we developed the synchronization tomography [2]. First the cerebral current source density is estimated by means of magnetic field tomography (MFT) [3] for each time step; then the single-run phase synchronization analysis is applied to the current source density in each voxel of the reconstruction space. We model different generators of ongoing rhythmic cerebral activity by current dipoles with time courses of slightly detuned coupled chaotic oscillators subjected to random forces. MEG signals are calculated using these generators for a whole-head MEG system, and subsequently the synchronization tomography is applied to the simulated measurements. We show advantages and limitations of the synchronization tomography. We demonstrate that the measurement noise resistance of the synchronization tomography is much higher compared to MFT alone. Furthermore, we compare different approaches to extend phase synchronization analysis, which was developed for scalar signals, on vector valued signals like the current density. The localization accuracy can be considerably improved using the strongest component of the principal component analysis of the reconstructed current density vector. We applied statistical tests using surrogate analysis [4] to test synchronization tomography for its characteristics concerning non-linearities of sources. In other words, we test whether synchronization tomography removes non-linearities of sources or introduces non-linearities not present in the sources.

References: [1] Tass PA et al. Phys Rev. Lett 1998; 22: 3291–3294. [2] Tass PA et al. Phys Rev Lett 2003; 90: 088101. [3] Ioannides AA. Inverse Problems 1990; 6: 523–542. [4] Dolan K, Neiman A. Phys Rev E 2002; 65: 026108.