A model-based detector of vertex waves and K complexes in sleep electroencephalogram.

A model of sleep phasic events such as vertex waves, K complexes, delta waves and sleep spindles is proposed. It consists of feedback loops that are driven by white noise (simulating tonic delta and sigma activity) and by isolated random impulses, simulating vertex waves or K complexes, depending on the background tonic activity. A model-based method for the detection of sleep phasic events was implemented in a personal computer. Its performance was investigated using simulated and real whole-night EEG signals. The method was able to detect K complexes and vertex waves in a reliable way in spite of their variable shapes and in the presence of a variety of background activities. The detector appears to have superior performance to those so far reported in the literature. The performance of the detector was also compared to that of an electroencephalographer using normal sleep EEG records of 8 h duration from 6 subjects. The performance was satisfactory both in terms of accuracy and reliability. The problem of detecting K complexes in stages 3 and 4 of sleep is discussed.