engleski

Spatio-temporal analysis of simulated neuromagnetic responses to faces

Magnetoencephalographic (MEG) recordings and various spatio-temporal source localization methods are often used to study cortical responses to faces. Face processing involves activation of multiple cortical areas often with a pronounced synchronicity which can adversely affect spatio-temporal source parameter estimation accuracy. The goal of our simulation study was to explore conditions which allow identification of multiple synchronously active focal sources in particular as a function of the width of the selected temporal window. The simulated data sets were generated using a forward simulator in the MRIVIEW software package (Ranken et al., 2004). The segmented cortical surface from subject's MRI was used to determine the local cortical geometry. Locations, strengths, and dynamics of the simulated point current dipole sources were defined to correspond to the typical findings from our empirical data (Susac et al, 2004, 2008) ; the source orientation was assumed to be perpendicular to the local cortical geometry. A Gaussian white noise was added to the simulated MEG responses to calculate noisy forward fields, which were then submitted to the Calibrated Start Spatio-Temporal (CSST) inverse procedure. The semi-automated CSST algorithm randomly selected initial starting locations from the previously segmented cortical surface and thousands of unconstrained independent fits in a search for a global minimum, using data from all sensors. Our results revealed that it was possible to identify multiple active sources (e.g. up to seven) when the whole time interval was analyzed and thousands of the initial starting locations were used. However, locations and dynamics of the identified sources were accurate only for certain source configurations in which close sources were not simultaneously active. Our simulations demonstrated improved source parameter estimation accuracy when using smaller time intervals, especially when sources were rather close, as often encountered in the analysis of neuromagnetic responses to faces.

MEG; numerical simulations; temporal dynamics; source models

nije evidentirano