Predicting sleep stages with machine learning and wearable byteflies sensor dots: a pilot study
Carroll, James Peter
MetadataShow full item record
The conventional method for quantifying sleep is through the use of Polysomnography (PSG) and a trained human sleep scorer by observing and evaluating the output in 30-second epochs. A PSG device can be rather invasive to one’s regular sleep pattern and therefore can potentially result in irregular sleep patterns. Furthermore, human sleep scoring classification by a trained expert can be rather time consuming and subject to inter/intra rater variability. Nevertheless, human sleep scoring with PSG still remains the gold-standard for sleep measuring and classification for the diagnosis disorders related to sleep. The present pilot study explores the possibility of using a wearable device known as a ByteFlies Sensor Dot to measure signal activity from an individual during a night’s sleep. This validation study focuses on the signal capture of alpha frequency band through a phenomenon known as “the Berger effect.” Participants will be asked to open and close their eyes while being connected to the gold standard PSG device and exploratory ByteFlies Sensor Dot device. The resulting alpha signals will be identified with a machine learning algorithm for cross comparison and analysis. In conclusion, the validation study will discuss methods to improve on the measuring of EEG and sleep stage scoring with the ByteFlies Sensor Dot for sleep monitoring and sleep disorder diagnosis.