Performance of Automated Detection Software in Absence Seizures

Absence seizures are common in children with idiopathic generalized epilepsies. Identifying absence seizures in EEG using automated detection software offers potential efficiency gain.

This study assessed Persyst's performance in detecting absence seizures (≥3 seconds) in prolonged pediatric EEGs. Two detection modules were used: seizure detection (most sensitive settings) and spikeburst detection which detects 'bursts' of interictal epileptiform discharges (≥1 spike per second) with a minimal duration of 1, 2 or 3 seconds.

Eighty-one EEG records from 69 patients were analyzed. Seizure detections sensitivity is 78%, spikebursts sensitivity is 95% (1 second), 77% (2 seconds) and 61% (3 seconds). Detection rate of both modules increased to 99% for absences lasting more than 9 seconds. Median false positive rates are 24 per 24 hours (range 0-60) for seizure detection and 23 (range 0-293), one (range 0-36) and zero (range 0-19) for the spikeburst module with duration of respectively 1, 2 or 3 seconds, primarily due to interictal discharges.

Spikeburst module with a threshold duration of 2 seconds represents the most feasible compromise between sensitivity and clinical usability.