A State-Space Modeling Approach for Localization of Focal Current Sources From MEG

Abstract

State-space modeling is a promising approach for current source reconstruction from magnetoencephalography (MEG) because it constrains the spatiotemporal behavior of inverse solutions in a flexible manner. However, state-space model-based source localization research remains underdeveloped; extraction of spatially focal current sources and handling of the high dimensionality of the distributed source model remain problematic. In this study, we propose a novel state-space model-based method that resolves these problems, extending our previous source localization method to include a temporal constraint by state-space modeling. To enable focal current reconstruction, we account for spatially inhomogeneous temporal dynamics by introducing dynamics model parameters that differ for each cortical position. The model parameters and the intensity of the current sources are jointly estimated according to a bayesian framework. We circumvent the high dimensionality of the problem by assuming prior distributions of the model parameters to reduce the sensitivity to unmodeled components, and by adopting variational bayesian inference to reduce the computational cost. Through simulation experiments and application to real MEG data, we have confirmed that our proposed method successfully reconstructs focal current activities, which evolve with their temporal dynamics.

Effects of reconstructed magnetic field from sparse noisy boundary measurements on localization of active neural source

“…(8) converges to that computed from the forward model Eq. (1) provided the number of expansion terms M l for degree l is sufficiently large.…”

“…• For measurements with low SNR, Methods 2a and 2b (that bases on reconstructed data) offer a significantly more accurate localization than Method 1 (where direct measurements were used) despite errors in data interpolation and field reconstruction. This is because the BC integrals over the entire surface (instead of scattered interpolated data) are taken into account in the physics-based model (8) to reconstruct a smooth MFD leading to a more representative set of "reconstructed measurements" for ANS location. …”

“…(8) and (9) is reconstructed. Since interpolation and reconstruction introduce errors in the presence of measurement noise, a gradient-based method is developed to establish a rational criterion for selecting high-fidelity reconstructed data to improve ANS localization.…”

“…(8). At any specified location, the reconstructed B r can be viewed to consist of two parts, B c and B n , as shown in (14c), corresponding to noisefree and only noise measurements B si, respectively.…”

“…Herein, nonlinear parametric models are formulated as an optimization problem (that minimizes the difference between measurements and model predictions) and solved with a least-square (LS) method for the six parameters required to characterize the position vector L and dipole moment Q of the ECD [11,12]. In ANS localization studies, the extremely weak MFD measured outside of the head is several orders smaller than its external interference; for instance, the power ratio of MEG signals to noise equaling 1 is common for clinical event-related measurements [7,8,10].…”