Automated Sleep Spindle Detection System using Period-Amplitude Analysis

Abstract

Sleep spindles are rhythmic transient waveforms present in the electroencephalogram (EEG) of non-rapid eye movement (NREM) sleep. In the present study a period-amplitude analysis method was applied for the automated detection of sleep spindles in all-night sleep EEG recordings of young healthy subjects. The method relies on the characterization of individual half-waves of the EEG data, by estimating electrographic parameters such as amplitude and duration and by assigning a grade to each half-wave depending on where it lies in the amplitude-frequency plane. The grading is followed by the detection system, checking consecutive half-wave characteristics and implementing a set of rules for determining the start and the end of spindle bursts and for retaining or rejecting sleep spindle indications provided during the various stages of the detection system. The sensitivity and false positive rate across subjects was 78.9% and 10.9%, respectively, providing indication that the method could be successfully applied to larger sets of healthy subjects of various age groups, as well as to patient populations.