Detecting Sleep Spindles Using Entropy

Sleep spindles are bursts of brain activity during sleep. They occur during the NREM2 stage of sleep and appear as fluctuations in electric recordings, looking like yarn spindles. This increase of activity can be detected by complexity measures, the most popular of which are the entropy based estimations. In this paper, we use entropy to measure the brain activity during sleep spindle and non-spindle periods and discriminate them employing the machine learning technology. Two are the main outcomes of this work: a) we show that it is possible to achieve remarkable classification performance when detecting sleep spindles with entropy based measures and machine learning techniques, presenting classification accuracy of more than 95\(\%\) and (b) we report that bubble entropy, a recently introduced definition of entropy, presented the lowest p-value of all examined features.