engleski

Optimal spatio-temporal matrix subdivision for cortical neurodynamics estimation

OBJECTIVE: Magnetoencefalography offers a unique insight into the cortical neurodynamics at a millisecond time scale. Its estimation from the averaged data however, can be affected by source modeling assumptions and estimation procedures and strategies. We conducted numerical simulation studies to examine the accuracy of the neuromagnetic time course estimation depending on the choice of the spatio-temporal data matrix and the source location biases caused by the extent of the active cortical sources. MATERIALS & METHODS: Numerical simulations were conducted using the MRIVIEW Cortical Start Spatial Temporal (CSST) multi-dipole localization program. Up to 7 point current dipoles and/or extended patches of cortical activity were placed on a real cortical surface with a wide range of temporal corellation patterns. From 1000 up to 50 000 starting points restricted to the realistic cortical surface within a sphere were randomly selected during a two-stage multi-start simplex minimization procedure. The time intervals considered included a range of regular time windows as well as data guided choices. RESULTS: For long time intervals in which 5 to 7 current sources were activated, a subdivision into smaller time windows was necessary for most of the configurations with prominent temporal corellations. Contour maps guided selection of the time windows was useful and resulted in accurate time course estimations. The selection of regular time intervals in which the spatio-temporal data matrix was subdivided, independent from the starting latency, allowed also an accurate time course estimation but only for a limited range of time intervals. Inaccurate and/or less efficient time course estimation was obtained for smaller and larger intervals, respectively. The optimal range was found source configuration dependent. The effects of source extent on the time course estimation were less pronounced and only in some cases affected the onset and offset of activity. CONCLUSION: Our results suggest that a more efficient and accurate time course estimation could be obtained by identifying the optimal regular time interval width in which the spatio-temporal data matrix could be subdivided. The identification of such an interval required a rather coarse search suggesting that an automated procedure could be performed in the future.

MEG; temporal dynamics estimation; spatio-temporal matrix; numerical simulations

nije evidentirano