Automatic recognition of spike and wave bursts.

This paper reviews previous methods and describes a new method for the detection of spike and wave bursts in a single EEG channel. The complete detector is readily implemented in a 16-bit microprocessor. The burst detection relies on the detection of both spikes and waves and also imposes a requirement on the repetition rate of the slow waves. The system agreement with electroencephalographers was over 90% for spike and wave bursts longer than 3 sec in duration, with a false detection rate of one error in 5.8 h for bursts of the same length. The system also computes the mean value and variance of the amplitude and duration of the spikes and slow waves as well as the period of the spike and wave complexes. The EEGs of 6 patients suffering from petit mal epilepsy were analyzed. The variance of the amplitude measures was found to be much greater than that of the period measurements. For all patients the repetition period of the waves decreased for longer bursts, as did the variability of the repetition period. The difficult problem of false detections due to artifacts is discussed in detail.