Reliability of automatic and visual analysis of interictal spikes in lateralising an epileptic focus during video-EEG monitoring

Abstract

We evaluated 2 h of two night recordings of surface EEG of 10 patients with drug-resistant focal epilepsy using video-EEG monitoring, giving 40 h of EEG. The raw data of the automatic spike analysis according to the Gotman algorithm was visually corrected by rejecting false detections. Furthermore, the complete EEG recordings were analysed independently visually by two experienced electroencephalographers. For each method we analysed the total count of detections and the topographical distribution (left-right) of spikes. The total number of detections was significantly higher (243%) in the raw data and significantly lower after elimination of false detections (57%) in comparison to conventional analysis (100%). Lateralisation was concordant between the methods in 9 10 patients. The extent (<75%, 75–90%, ≥90%) was concordant in 80% between the two human raters. The automatic analysis with elimination of false detections was concordant with each of the human raters in 60% of patients. Extent of concordance was dependent of the total number of spikes with patients having more spikes being more reliably lateralised. Our results suggest that visually corrected automatic spike analysis is an economical method to use interictal epileptogenic activity as an independent indicator of the side of the epileptogenic focus in the setting of non-invasive presurgical evaluation. This is especially true in patients with many spikes.