Adaptive Array-Gain Spatial Filtering in Magnetoencephalography
Thomas Maloney · OhioLink ETD Center (Ohio Library and Information Network) · 2010
This paper analyzes adaptive array-gain spatial filtering as a way of interpreting magnetoencephalography (MEG) data to locate centers of neuronal activity within the brain.When neurons in the brain communicate they generate an electric current which, in turn, generates a magnetic field.MEG can record this magnetic signal when it is generated simultaneously in multiple neurons.The features of MEG include good spatial resolution, very good temporal resolution, and no radiation.Currently, inverse modeling is used in the clinical setting to interpret the MEG signal.Since there is no unique solution to the inverse method, the results rely on the skilled interpretation of a neurologist to evaluate their concordance with the patient's clinical symptoms.Spatial filtering is a way of locating sources of power within the brain that avoids having to use inverse methods and, therefore, avoids many of the localization errors and artifacts.Spatial filtering accomplishes this by using the relationship between the signals from the individual detectors, in the form of a covariance matrix, along with the known detector sensitivity to magnetic dipoles, known as the lead field.In this study, the head is modeled as a homogonoues conducting sphere and the neuronal activity is modeled as an electrical current dipole.Both patient data and simulated data are analyzed in this study.For the simulated data, dipoles at various positions and orientations are analyzed with and without various levels of Gaussian white noise added in.The results of this study show that adaptive array-gain spatial filtering has the potential to become a standard method in the localization of neuronal activity centers in the brain.With its propensity for reduced error and speed in interpreting the data, spatial filtering may one day help to make MEG a standard procedure in pre-surgical brain mapping, thereby reducing patient morbidity.iii This paper would not have been possible without the help of the following people.Dr. Douglas Rose and Nat Hemasilpin of Cincinnati Children's Hospital Medical Center for supplying me with the patient data used in my analysis.Dr. Rose also introduced me to MEG and gave much support and guidance and Nat helped me work with the data in Matlab.Dr. Stephen Robinson of the Henry Ford Hospital for his help on some of the technical aspects of spatial filtering.Dr.