Some characteristics of sleep spindles derived from automatic analysis.

Sleep spindles in eight normal adults were analyzed automatically by a minicomputer. The accuracy of spindle detection was about 90%. Spindle wave frequency was measured by an approximation method for the spindle peaks; this method employed a quadratic equation on prior analog filtering. The accuracy of measurement was within 0.6 msec (SD 1). The average spindle wave frequency was about 13 Hz, but varied depending on the sleep stage and brain region. The frequency was highly variable and displayed no tendency. Concerning spindle duration, short spindles were followed by either short or long spindles, while long spindles rarely were followed by long ones. As sleep progressed, spindle duration tended to become shorter. There was a tendency for one long interval of spindle appearance not to be followed by another long interval of appearance. However, as sleep progressed, the interval of spindle appearance had a tendency to become prolonged.