Statistical Signal Processing Method for Sleep Analysis in Infants

Abstract: Sleep is a natural periodic state of rest for mind and body taking about one third of lifetime of humans. It is essential for their well-being, efficiency and long-term health. During infancy and early childhood, the role of sleep is even more crucial as human infants spend more time asleep than awake and optimal conditions in this stage are necessary for their physiological, mental and social development. This has stimulated numerous conventional researches about infant sleep which have investigated qualitatively the role of sleep on infant development. However, such studies have not included experimental manipulations designed to evaluate the underlying relation between the disposable diaper, as one of multiple factors affecting the sleep of infants, and the quality of sleep.

In this thesis, a new method for monitoring and quantitatively analyzing the sleep of infants was developed; depending only on the body movement signal measured usin an accelerometer. For this, a sleep study involving 36 healthy infants aged 28+/-6 weeks was performed over 18 nights. The analysis of the activity signals was carried out by an in-depth signal processing method. To support the preprocessing of the measured data, daily sleep diaries and questionnaire were also used. After extracting multiple local and global features from the processed signals, the output is then used to build a randomized comparative analysis. Its main purpose is to evaluate quantitatively if there is an influence of using either a weakly or highly absorbent disposable diapers, as one of multiple factors on the quality of infant sleep and in which features of the activity signal variations can be observed.

The developed algorithm allows for detecting and analyzing high-activity peaks in the signal. The number of extracted peaks correlated highly with the number of awakenings reported in the sleep diaries. We also found that even though the sleep duration was comparable between the two groups of nights using different products, there were measurable variances in both the level of activity and the sleep fragmentation. The differences were observed in the global features of the signal, while no significant changes were detected when comparing the local features within the night. Furthermore, the study showed a higher level of activity in boys compared to girls, and a higher sleep fragmentation for breastfed infants compared to those who relied on formula milk for nutrition. Finally positive influences and limitations of the study are discussed.