Approaches to Inverse Problem Estimation

Richard Leahy
University of Southern California
Signal and Image Processing Institute

Magnetoencephalography (MEG) is a method for imaging function in the human brain that measures the magnetic fields produced by neuronal current flow. As a result of their local alignment and
morphology, pyramidal cells in cerebral cortex produce magnetic fields outside the human head as a result of normal electrical brain activity. These fields are very small and can only be
measured with specialized SQUID magnetometers. MEG involves the study of brain activity using these devices. MEG measures neural activity directly, rather than indirectly as in the case of fMRI, so that temporal resolution can be far higher (~1ms or higher) than with fMRI (~1sec or lower) and we can study the dynamics of brain activation. MEG has wide application in studies of sensory, motor, and cognitive activity in the cerebral cortex. Clinically, it is useful for presurgical mapping of sensory and motor cortex
and, in the case of epilepsy, for localizing potential sites of abnormal activity that may cause seizures. I will describe how we use MEG data to estimate the locations of neural activation either as a set of discrete focal sources or as sources distributed over
the cerebral cortex. Challenges in source estimation include solving the associated forward problem, dealing with hugely under-determined systems of equations, and assessing the
reliability of the estimated sources. I will address each of these problems to provide an overview of the issues involved in analysis
of MEG data and the potential of the modality for providing unique insights into human brain function.

Presentation (PDF File)

Back to Long Programs