Abstracts

Rationale: The responsive neurostimulator (Neuropace RNS^® System, Mountain View, CA, USA) provides closed-loop detection and electrical stimulation with intracranial electrodes to the seizure focus. The purpose of this study was to quantify the interrater reliability of electrographic seizure (ESz) classification in electrocortigraphy (ECoG) samples recorded by the neurostimulator. Methods: ECoGs were recorded as part of a randomized controlled trial (Morrell et al, Neurology 2011). The ECoGs used in this analysis comprised either a "long episode" (detection continued beyond physician-defined duration) or a "saturation" (high amplitude activity). Board-certified/eligible reviewers interpreted ECoGs while blinded to neurostimulator settings, patient/electrode characteristics, and other interpretations. We defined an ESz as a sustained rhythmic discharge, including repetitive spiking or spike-and-wave at ≥ 2 Hz, with definite evolution in frequency, location, or morphology, and clearly distinguishable from background, lasting at least 10 seconds. Reviewers classified each sample as "seizure" or "not seizure" (epileptiform activity or nondescript pattern). An agreement rate and Fleiss' kappa were calculated overall and for individual pairs of reviewers. ECoGs formed 3 groups by interrater results: 1) those in which both reviewers identified as ESz; 2) one reviewer ranked as ESz and the other as "non-seizure"; and 3) both reviewers identified as non-seizures. Ten examples from each group were then randomly selected for further characterization. Results: Five reviewers forming 6 pairs reviewed a total of 4280 ECoGs from 131 patients. The overall agreement rate (both reviewers of each pair agreed ESz or "not seizure") was 77% for a kappa = 0.541 (moderate agreement). Agreement rates between pairs of reviewers ranged from 0.67 to 0.93. The range of kappa statistics between pairs was from 0.42 (moderate) to 0.85 (excellent). The ten ECoGs with agreement for ESz clearly fulfilled judging criteria. Although the ten "nonseizures" did not qualify as ESz, all contained abnormal epileptiform activity. The ten ECoG samples with mixed interpretations fell into three categories: 1) low amplitude rhythmic activity with limited spatial or temporal evolution until the end of the discharge; 2) discharges with clear evolving activity that were less than 10s because ictal onsets - and thus, the overall duration - were ambiguous (some overlap with the first category was present); and 3) quasiperiodic discharges during which runs became more organized with frequencies near 2Hz. Conclusions: The majority of the ECoG detections were classified by experts as ESz with moderate interrater reliability. Other events were bursts of activity that did not qualify as a ESz as defined by the study but were clearly abnormal and epileptiform. Disagreement in classification may have origin in the somewhat arbitrary but necessary definition of ESz. Future work may correlate clinical effectiveness with the type of neurostimulator detection and guide definitions of intracranial ESz and their detection.