No abstract

No abstract

A discrimination between the two different types of acute stroke (ischemic and hemorrhagic) is accomplished by the implementation of both the Inverse problem of electroencephalography (EEG) technique and the method of principal components analysis (PCA). The study was based on electroencephalograms (EEGs) recorded from patients that had suffered from stokes. The brain activity was simulated with a realistic head model excited by electric dipoles, which in the present work were allowed only to rotate about a fixed origin. Combining the calculated surface potentials of the head model and the EEG recordings, the inverse problem algorithm converged to a solution giving an equivalent dipole that included all the information needed to distinguish each type of stroke. Alternatively, PCA technique was implemented directly on the EEG recordings in order to reveal potentially hidden patterns underlying the recordings. For this purpose, the corresponding techniques developed within our previous work, are exploited herein for the processing of patients' EEGs. It is observed that indeed both equivalent dipole and PCA or its alternative proper orthogonal decomposition approaches were able to discriminate the two types of stroke.

No abstract

A method for the equivalent brain source localization from EEG measured data is proposed herein exploiting the combined advantages of Proper Orthogonal Decomposition (POD) and Finite Element Method (FEM). The current effort is concentrated on solving a fine discretized forward eigenproblem once and applying POD to reduce the problem dimension. The equivalent brain source localization is in turn formulated by setting up a cost function in its least squares means. The involved calculated data sets as well as the sensitivity matrix are evaluated herein exploiting a POD eigenfunction expansion. Explicitly, for each forward problem solution the voltage within a small volume enclosing the source is expressed using FEM, while for the remaining larger volume a POD eigenfunction is adopted. The continuity of the two solutions is enforced following a Dirichlet-to-Neumann map formalism. The proposed method is expected to offer a computationally efficient forward and overally inverse problem solution.