Methods for detecting high-frequency oscillations in ongoing brain signals: Application to the determination of epileptic seizure onset zones

METHODS FOR DETECTING HIGH-FREQUENCY OSCILLATIONS IN ONGOING BRAIN SIGNALS: APPLICATION TO THE DETERMINATION OF EPILEPTIC SEIZURE ONSET ZONES Chiran D. Doshi, B.S. Marquette University, 2011 Epilepsy is a neurological disorder with varied expression. Patients with focal onset seizures that are resistant to medications can benefit from ablative surgery. However, localization of the seizure onset zone (SOZ) and characterization of propagation to secondary areas can be challenging. The present study is aimed at developing appropriate signal processing methodology to detect bursts of interictal highfrequency oscillations (HFOs), as a possible signature of the SOZ, in patients with drugresistant partial epilepsy. Invasive interictal and ictal intracranial electroencephalography (iEEG) data and non-invasive electromagnetic source imaging with magnetoencephalography (MEG) data from three subjects were analyzed. We developed a novel algorithm that extracts HFO bursts from the envelope of iEEG and MEG traces in the [80-300] Hz range. Clusters of HFO events were subsequently analyzed to investigate their relative time delays and to infer possible propagation patterns during the interictal period and episodes of ictal onset (iEEG only). The location of iEEG electrodes were labeled with respect to the chronometry of the detected HFOs. The recording site bearing the smallest rank was labeled as the lead generator of HFO discharges. The aim of using MEG traces was essentially to determine probable SOZ locations. We proposed a new metric referred to as ‘spiking index’ that was computed at each cortical site in the vicinity of iEEG electrode locations (iEEG and MEG data were obtained for the same patients: iEEG was considered as the standard of reference for MEG results). The sensitivity and specificity of the HFO detector operating from ongoing brain traces were evaluated. Our results indicate that higher values of spiking index and higher rates of HFOs corresponded to brain regions that were identified independently as the SOZ by an expert clinician and as determined by the location and extent of the cortical resection that freed the patients from the seizures. Interictal and ictal iEEG HFO localization showed good concordance with the location of resected areas. The use of interictal data only, if used for surgical planning, would reduce the time required for making decisions regarding the resection of cortex and improve the chances of successful surgery. Obtaining iEEG data is invasive, with possible risks to the patients. Another fundamental disadvantage of iEEG is that the implanted electrode grids and strips needed to cover the supposed abnormal cortical areas for proper determination of the SOZ. Our results indicate that the spiking index and rate map obtained from MEG source maps may provide a non-invasive alternative for determination of the SOZ and may provide greater accuracy to the placement of the implantable electrodes, and eventually avoid an invasive exploratory procedure before surgery.