Automatic classification of sleep stages for children using a wearable device

Details

Supervisors

Verleysen, Michel ; Lee, John Aldo

Faculty

Ecole polytechnique de Louvain

Degree label

Master [120] : ingénieur civil biomédical, à finalité spécialisée

Abstract

Sleep related breathing disorders (SRBD) are known to have adverse effects on infants. Indeed, it was proven to be strongly linked with abnormal cognitive development and Sudden Infant Death Syndrome. Therefore, it is of prime importance to diagnose SRBD as early as possible to intervene and ensure a normal and healthy development of the patient. The current golden standard for assessing the quality of sleep of an infant is Polysomnography (PSG). It consists in a very complete exam (EEG, ECG, EMG, Air flow measurement...) performed in a special facility or unit of the hospital overnight. PSG has a number of drawbacks : it is inconvenient for the patient and his family, it requires the presence of trained technicians to equip the patient, check that the procedure goes without failures and interpret the recordings. Moreover, the hospital environment may change the infant’s behaviour and therefore have an impact on his sleep during the recording. An alternative is at-home monitoring, using an armband. Gabi SmartCare has developed such a band as well as a whole processing system downstream in order to detect suspicious events and diagnose SRBD without the need of a PSG. A part of this process is the classification of the recording clips in sleep stages, as the sleep stage in which an event happens affects the interpretation that is made of it. The aim of this thesis is to implement that classifier. We show that a classification can be performed using a pool of several patients as training and following the same process that Boe & al. used for a similar project on adults patients.