Brain Electromagnetic Source Imaging: 'Sourcery' or a Window on the Mind?

摘要

Quantitative inverse localization of the current sources responsible for scalp-recorded electric (EEG) and magnetic (MEG) brain waves has been attempted with varying degrees of sophistication and success in medical and neuroscience research applications. Following early observations that scalp fields were often dipolar in appearance, the first methods used idealized spherical head models and best-fit single equivalent dipole sources were estimated at each time instant from the field data using standard optimization methods. Useful results were forthcoming in situations when such restricted source assumptions were reasonable and absolute localization accuracy was not important. Since then, more realistic head models (boundary and finite element) have come into use and it is now possible to use a custom head model based on MRI or CT brain imaging. Source models are also potentially more realistic and allowing multiple, extended sources with a wide variety of anatomical and spatiotemporal constraints. Similarly, more robust methods of optimization are now in common use. This paper reviews the assumptions and basic limitations underlying the various brain source localization approaches and the degree to which they can provide useful information or illusory results. Particular attention is paid to noise limitations and the inherent insensitivity of scalp fields to detailed source characteristics, and the need for more sophisticated error analysis and validation studies.