Simultaneous estimation and testing of sources in multiple MEG data sets

• Published in: IEEE transactions on signal processing 2005-09, Vol.53 (9), p.3449-3460
• Main Authors: Bijma, F., de Munck, J.C., Huizenga, H.M., Heethaar, R.M., Nehorai, A.
• Format: Article
• Language: English
• Subjects: Applied sciences; Brain modeling; Component model; Confidence; Confidence intervals; constrained CramÈr-Rao Bound; coupled dipole model; Cramer-Rao bounds; Detection, estimation, filtering, equalization, prediction; Dipoles; Electroencephalography; Exact sciences and technology; Information, signal and communications theory; Joining; Magnetic analysis; MEG; Principal component analysis; Shape measurement; Signal and communications theory; Signal, noise; spatiotemporal covariance; Spatiotemporal phenomena; Telecommunications and information theory; Temporal logic; Testing; Time functions; Time measurement; trilinear model; Yield estimation
• ISSN: 1053-587X; 1941-0476
• DOI: 10.1109/TSP.2005.853097

The proposed Extended Couple Dipole Model (ECDM) is a trilinear component model that can be used to analyze multiple, related MEG data sets simultaneously. Related MEG data sets are data sets that contain activity of the same sources or activity of sources that have proportional source amplitudes. The simultaneous model uses a set of common sources and a set of common source time functions (wave shapes) to model the measured data in each data set. The set of common sources contains all sources that are active in at least one of the data sets to be analyzed. The number of common spatial and temporal components is specified by the user. The model for each data set is a linear combination of these common spatial and temporal components. This linear combination is estimated in a coupling matrix. Unlike the Coupled Dipole Model, where the user selects certain entries of the coupling matrix to be zero, the entire coupling matrix is estimated in the ECDM. This yields a more objective and statistically transparent estimation method, of which the identifiability constraints do not depend on the user's chosen design as in the CDM. Crame/spl grave/r-Rao Bounds are derived for the ECDM, and the significance of the estimated source activity is computed and illustrated by confidence regions around estimated source time functions.