Abstracts

Rationale: Epilepsy surgery is an effective treatment for patients with drug-resistant focal epilepsy, but up to half of patients continue to have seizures postoperatively. Suboptimal outcomes may result from inaccurate localization of the seizure focus due to electrode sampling bias or failure to record typical seizure events. Interictal epileptiform discharges (IEDs) occur more frequently than seizures but are non-specific, occurring over broader cortical areas than seizure onsets. This study presents a novel localization method based on the hypothesis that IEDs reflect the receipt of signal propagating either along the pial surface (GM) or through white matter (WM) from a focal source.

Methods: We studied 38 patients with drug-resistant focal epilepsy (25 male, age 35±11 years) who underwent intracranial EEG (iEEG) monitoring for epilepsy surgery evaluation at the National Institutes of Health Clinical Center. We acquired 3T T1w MRI and diffusion-weighted imaging for all patients. One year outcome data were available in 34 patients. Cortical surfaces were reconstructed, and streamlines were mapped to the Schaefer 600 atlas. We estimated WM bundle lengths and GM geodesic distances along the pial surface between each iEEG electrode and every cortical surface node. iEEG IED sequences were detected and clustered by sequence similarity. We identified most likely sources of IED sequence clusters using time-difference of arrival multilateration based on GM and WM distances and conduction velocities (Figure 1).

Results: Source localization resulted in a parcel distinct from the most frequent leading electrode in 49/67 clusters (73%). Localization using combined GM and WM propagation accurately predicted seizure outcome in 75% of patients; localization with GM propagation alone resulted in an accuracy of 72%; the leading electrode method resulted in an accuracy of 63%. Among sequences requiring WM propagation (+WM), 21.8% involved propagation to a distant electrode (farther than 99% of sequences using GM propagation) (Figure 2A). While the leading electrode tended to be closest to the source in sequences involving GM propagation, this was not the case in sequences requiring WM propagation where the closest electrode spiked in all sequence indices (Figure 2B).

Conclusions: Using a novel localization algorithm, we demonstrate that a focal source can explain broadly distributed spikes observed in IEDs and can be identified in the absence of iEEG coverage of the source region. Incorporating both GM and WM propagation increases the proportion of sequences explained by the model and improves the accuracy of seizure outcome predictions. Because WM conduction velocities are greater than that of GM, electrodes farther from the source receiving signal via WM pathways often spike before closer electrodes. Use of the leading electrode for IED source localization may therefore mislead clinicians in cases with significant WM propagation of IEDs.

Funding: NIH Intramural Research Program