Measuring Entropy in Sleep EEG to Examine Complexity and Level of Biological Activity in Different Sleep Stages

Sleep scoring is performed to help diagnose sleep disorders. It is a difficult and tedious process, usually taking several hours. However, by identifying characteristics that can be fed into a computer to identify sleep stages, it is possible to automate the process. Sleep scoring is performed using an electroencephalogram (EEG) to measure a person's brain waves during sleep. The data obtained from this reading can help split a person's sleep into different sleep stages. One of the characteristics that can be used to classify sleep stages is entropy. Entropy in biological systems can be thought of as a measure of disorder. It is often correlated with complexity in biological systems, in which higher entropy often means a higher level of complexity. We examine this phenomenon regarding sleep stage wakefulness, N1, N2, N3, N4, and REM, to identify whether a deeper level of sleep correlates with a lower level of entropy. A large publicly available EEG database of sleep datasets was used for this study. For each dataset, we calculated the approximate entropy and multiscale entropy. Results show a decrease in entropy as subjects fall further into a deeper sleep. For approximate entropy, entropy decreases from N1 to N4. A similar trend exists in MSE at lower time scales, but this trend is reversed at higher time scales. This may hint at a hidden complexity within deeper stages of sleep that may be related to unseen biological activity. Understanding the entropy traits inherent in sleep stages makes it possible to automate the process of sleep scoring, allowing for fast, accurate sleep scoring for patients at home.