Abstracts

Rationale: The single equivalent dipole model (ECD) is the standard for source estimation in clinical magnetoencephalography (MEG). While this approach is highly robust and statistically proven, there are a number of circumstances where the a priori assumptions required to obtain accurate source estimation are not met. MEG beamformer (BF) source localization algorithms provide a high temporal- and spatial-resolution source estimation of brain activity and have the special property of being robust in the presence of interfering noise. Methods: First, we evaluate the beamformer performance on simulated neuromagnetic source activity within real brain noise. Simulated sources were parameterized across levels of signal-to-noise-ratio (SNR) and multi-focal source/noise scenarios. The BF methods were compared to single ECD models for accuracy across at least 15 source samples per level. Secondly, we apply BF source localization combined with multivariate spatiotemporal decomposition methods to clinical case studies in which the standard ECD model fails in order to demonstrate the utility of the BF method in practice as applied to datasets that preclude the use of the single ECD model for functional mapping. Results: For the multi-source simulations event-related BF reconstruction accurately localizes transient activation at the various locations within the brain within the spatial sampling (+/- 5mm) of the reconstruction grid. Simulations varying signal source power show a consistent BF source localization depspite low SNR. Clinical case studies demonstrate how the BF can be used to mitigate sources of noise interference to allow localization of eloquent cortex in patient data, including hemisphere dominance for language, as validated against WADA and functional MRI. We also demonstrate the application of source level multivariate spatiotemporal decomposition to enhance our description of multifocal epileptiform discharges, reveal hyperactive epileptiform networks during quiescent resting state, and also provide an intuitive analysis of event-related language late-fields. Conclusions: While validation of the use of various distributed source localization models in clinical MEG continues, our results suggest that the MEG BF, used judiciously, has a unique and complimentary role in clinical MEG. Funding: none