Abstracts

Rationale: Epileptic spikes have been used as a surrogate marker of abnormal brain electric activity; however, spikes vary in morphology, distribution, duration, and amplitude. At times identification of individual spikes may vary among electroencephalographers. Using EEGs in children with Benign focal epilepsy of childhood with centrotemporal spikes (BECTS), we studied variability in spike identification between two electroencephalographers. Such spikes were chosen for analysis, as epileptic spikes in BECTS have characteristic morphology and distribution that are readily identifiable. Methods: This was part of a study on clinical and electrographic features of children with BECTS, approved by the institutional review board at Children s Hospital Boston. We analyzed routine awake and sleep EEG studies of randomly selected 5 children with BECTS (3 boys, 2 girls, age range: 7-11 years). For all studies, the first 5 minutes of stage II sleep were analyzed, where there were potentiation of centrotemporal spikes without significant contamination by movement or slow waves. Two blinded fully-trained electroencephalographers (Rater A and Rater B) marked each individual spike in the EEG segments, and a third blinded independent evaluator compared the results of spike markings. Spike identification by each rater was assessed by the evaluator, on a spike-by-spike basis, as well as comparing total number of identified spikes. Results: A total of 1500 seconds were analyzed over the 5 patients (300 seconds per patient). Rater A marked a total of 394 spikes (range 22-234 spikes per patient, mean 79 /-88 spikes), and Rater B marked a total 452 spikes (range 40-239 spikes per patient, mean 90 /-84 spikes). The difference in total of spikes between raters was 58 (13% of all spikes). For per patient, the difference between raters ranged 5-22 spikes per patient (2-50% of spikes, mean 12 /-7%). For spike-by spike analysis, discordance of identification between raters totaled 105 spikes (23% of all spikes). This ranged 13-27 spikes per patient, (5-55% of spikes, mean 37 /-19%). Major discordance was seen in spikes with less typical features for BECTS, such as 1) less typical morphology, 2) low amplitude, 3) less typical spread in distribution. Conclusions: Accurate identification of epileptic spikes remains a very challenging task. Even in BECTS, a relatively homogeneous epileptic syndrome with specific characteristics of epileptic spikes, our study found a great variability in identification of spikes. This was more emphasized when assessing on a spike-by-spike basis, while such difference became diluted when assessing total spike quantity only. Factors causing more variability in spike identification may need to be further delineated with future studies. Automated identification of individual epileptic spikes may be even less accurate compared to identification by trained electroencephalographers. Such factors associated with difficulty in accurate spike identification are significant for clinical care and research studies, as well as development of automated spike detection algorithms.