Abstracts

Rationale:
Quantitative analysis of intracranial EEG has the potential to add objectivity to the pre-surgical evaluation of candidates for surgical treatment of epilepsy. Several different quantitative biomarkers of the Epileptogenic Zone (EZ) exist. One such measure is the Epileptogenicity Index (EI), which yields high index values when low voltage fast activity (LVFA) exists at the onset of seizure. Another quantitative biomarker, called Slow Polarizing Shift index (SPS), calculates the area under the curve created when there is prominent DC shift at the onset of seizure. In this study we analyzed the performance of these two biomarkers individually and in combination.
Method:
We used clinical data from intracranial EEG recordings from 22 pediatric and adult patients evaluated at our institution. Using previously described methods, we calculated EI and SPS for each channel and up to 10 seizures for each of the 22 patients. We then performed ranked-sum comparisons of the median values of the individual indices and for the average of the two indices for channels that were inside and outside the Seizure Onset Zone (SOZ), which is the collection of EEG channels determined to be involved in the initiation of seizures. We also constructed Receiver Operating Characteristic curves (ROC) for the ability of the three different indices (EI, SPS, and the average of EI and SPS) to distinguish electrode contacts inside and outside the SOZ. We then grouped seizures by pattern of onset to see whether the area under the ROC (AUC) improved when both indices were combined.
Results:
We analyzed 213 seizures, with 2975 unique electrode contacts, from 22 patients who underwent intracranial EEG monitoring consecutively at the University of Michigan between 2013 and 2016. Median index values were significantly different for channels inside and outside the SOZ for all three indices: 0.06 versus 0.009 for EI (p < 0.00001); 0.27 versus 0.19 for SPS (p < 00001); and 0.21 versus 0.12 for the combination index (p < 0.0001). ROC curves showed the greatest AUC for seizures with polyspikes followed by LVFA or DC shift with LFVA. EI provide the best AUC in seizures with polyspikes followed by LVFA, DC shift followed by LVFA, and sharp waves in the theta/alpha range. SPS provided the best AUC for seizures with sharp waves in the beta range. The combination of the two indices produced the best AUC for seizures with LVFA at onset, preictal spiking followed by LVFA, and for slower rhythmic spikes less than theta range.