Abstracts

Rationale: About one-third of epilepsy patients do not respond to any medication, and surgical removal of the pathological tissue is a viable option to stop seizures. High-frequency oscillations (HFOs), ranging from 80 to 500 Hz, have been observed in both local field potential and scalp EEG recordings as potential biomarkers of the epileptogenic zone (EZ) to guide resective surgery and improve postsurgical outcomes. However, scalp HFOs, though promising in revealing the underlying epileptogenicity noninvasively, are impeded from clinical applications mainly by the factor that these low-amplitude and brief events are relatively difficult to be correctly distinguished and utilized in noisy scalp recordings. In the present study, we aim to image a subtype of HFOs, ripples riding on spikes (sRipples), via our proposed Spike Ripple Imaging Algorithm (SPIRAL) to noninvasively delineate the EZ for patients suffering from focal epilepsy. These events have been shown to occur more within the seizure onset zone (SOZ) than outside and may have great epileptogenic importance. Methods: We have collected and analyzed pre-surgical MRI and high-density EEG recordings from 10 medically intractable epilepsy patients who underwent surgery and became seizure-free after at least 6 months of follow up. Individual and realistic head model was made and on average 95 inter-ictal spikes were extracted for each patient. These spikes data were first pre-processed to remove bad channels/artifacts and examined and labelled via our in-house detector to find the putative sRipples. The detected events were then extracted as multichannel epochs from the noisy scalp recordings for further source imaging analysis via SPIRAL. We estimated the underlying spatial distribution of the sRipples for each patient and validated the results by comparing to the clinical evidence, surgical resection, which is modeled from the post-surgical MRI. For comparison, we also investigated the source imaging results of averaged spike for each patient.  Results: In this cohort of patients, the average localization error between estimated source and surgical resection is around 3.4 mm using sRipples comparing to 11.2 mm using interictal spikes. To evaluate the relative extent of the estimated sources, we computed the overlap between the estimated source and the ground truth (surgical resection) and normalized by the area of estimation and ground truth respectively to get the sensitivity and precision of the extent estimation; the geometric mean of these two values (termed as overlap rate) was used for validation. The overlap rate ranges from 0 (no overlap) to 1 (perfect match). Our proposed approach achieved an averaged overlap rate of 77% comparing to 54% of spike imaging, which indicates that sRipples in general have a better coverage of the surgical resection comparing to interictal spikes which are more spread in terms of cortical activation in our tested cohort. The preliminary results are summarized in Fig. 1. Conclusions: Our proposed approach to image and study sRipples, in a cohort of 10 focal epilepsy patients, demonstrated that the scalp sRipples can be adopted as a potential and effective biomarker to delineate the EZ noninvasively. The capability of SPIRAL to localize and determine the extent of the underlying epileptic brain sources could be of crucial interest in studying epileptic brain activities and making a potential impact on patient populations by providing guidance to the placement of the invasive electrode grids and additional insights to clinicians for the pre-surgical diagnosis. Funding: This work was supported in part by NIH R01 NS096761 and EB021027.