State dependent spike detection: validation

We present a formal validation of the method of state dependent spike detection; its principle and initial results were given in Gotman and Wang (1991). The method utilizes different procedures for elimination of non-epileptic transients according to state (active or quiet wakefulness, sleep stages); it also varies detection sensitivity according to state. Twenty new 100 min recordings were obtained from 20 unselected patients, covering active wakefulness, quiet wakefulness and various sleep stages. Results indicated that, while the total number of detections was reduced by 15%, true detections increased by 39% and false detections decreased by 60%. This significant improvement is due to the fact that wave morphology is now considered within its spatial context and particularly within a wide temporal context.