Reconstruction of intracranial evoked current distribution images using generalised matrix inverses and algebraic reconstruction techniques

Inversion of event-related potentials (ERPs) is studied using simulated potentials, through an analytic technique providing information about extended intracranial distributions, with separate source and sink positions. Comparative performance evaluation of the generalized matrix inverse (GMI) technique and algebraic reconstruction technique (ART) shows that for matrix dimensions used in inverse electroencephalography direct matrix inversion is faster than ART, in the noiseless case. Both methods present the same limitations concerning radial resolution due to their property of providing minimum norm solutions. When noise is present ART algorithms specifically designed for noisy data converge better to the original simulated current source distributions than GMI solutions, even if truncated approximations of pseudoinverses are used