Source-Estimation from Non-invasive Recordings of Brain Electrical Activity in Sleep and Epilepsy

Abstract

Epilepsy and sleep are characterized by spontaneously occurring large graphoelements in electroencephalography (EEG) and magnetoencephalography (MEG) recordings, such as ictal and interictal epileptiform discharges in epilepsy and K-complexes (KC) in sleep. Localization of the neural sources of these graphoelements, which is of immense clinical and research importance, requires application of electromagnetic source analysis methods. A number of such methods are available; however their ability to localize widespread synchronous cortical sources, such as the sources of KCs and widespread epileptiform discharges, is contested. Here, we used KC as an exemplar of large graphoelements with such sources to test the performance of a diverse set of commonly employed source analysis methods. We analyzed segments of sleep MEG data with clear KCs using equivalent current dipole models, beamformer methods, linear distributed source methods, and a non-linear distributed source method—magnetic field tomography (MFT). MFT provided the most robust and steady localization across KCs, which was also highly consistent with the intracranial findings: strong and widespread activations were reliably found in superior aspects of bilateral frontal cortex. Conversely, the localizations provided by the other methods were very variable across KCs and were all inconsistent with the intracranial findings: in many cases, the KCs were incorrectly localized in deep medial brain structures. Our current and earlier results showing the excellent localization accuracy of MFT for focal as well as extended brain sources and the smart uses of MEG and EEG in epilepsy, demonstrate that the MFT analysis of MEG signals may be a powerful tool for the studies of epilepsy, epilepsy monitoring and in pre-surgical evaluation of patients.