Error-reduction in dipole-source localization models

DESCRIPTION(Adapted from applicant's abstract): Multiple-channel brain evoked-response data is often recorded from the scalp, either electrically, or magnetically. The analysis of this multi-channel data commonly utilizes mathematical models, which assume that the neural generators of the evoked-response are dipoles. Such DSL can be accurate or inaccurate, depending upon many factors, most of which are not accessible to the researchers or clinicians who use the DSL. The problem is compounded by the fact that a given model may be accurate for dipoles in some locations, but not accurate for dipoles in other locations. Furthermore, the use of the least-square error by the DSL has been shown to increase dipole parameter errors. A method of evaluating whether insidious errors may be present in DSL analysis has been developed, based upon the fact that poor models show a sensitivity to neural-generator wave shape. The applicants have developed a proprietary, technically innovative method to utilize this sensitivity as an error-signal to allow adaptive-adjustment of model constants to reduce errors. The Phase I research will test this method using computer simulations. This research is health-related because DSL methods are widely used to analyze human brain activity, whether normal or abnormal. PROPOSED COMMERCIAL APPLICATION: The methodology will be incorporated into existing or future DSL software under license, since it would remove the present uncertainty of DSL results.