T-Pattern detection in physiological signals during sleep

H.G. Helgason², I. Hjálmarsson², G.K. Jonsson¹, M.S. Magnusson³, S. Kristjánsson²

¹Human Behavior Laboratory, University of Iceland, Univerisy of Aberdeen & PatternVision, Reykjavik, Iceland
²MedCare, Reykjavik, Iceland
³Human Behavior Laboratory, University of Iceland & PatternVision, Reykjavik, Iceland

Sleep recordings consist of various physiological signals, such as EEG, EMG, EKG, EOG, airflow, effort, and SpO2. Current analytical methods in sleep focus on counting specific events that are determined by predefined criteria and scoring rules. In addition, specific parasomnias are counted, such as apneas, hypopneas, desaturations and snoring periods, as well as periodic limb movement events to summarize the causes of sleep related problems. Indices are calculated from these events, diagnosis and treatment recommendations are given.

The main focus of our research is to search for physiologically based sleep patterns that have not been de.ned in advance and to explore the clinical application of the findings. We introduce a new approach, known as T-pattern detection, to the analysis of polysomnography data. Sleep research data was collected with Medcare Embla - Somnologica sleep diagnostic system and analyzed with Theme™ (PatternVision Ltd., Iceland).

The investigation highlights the potential for T-pattern analysis to make a significant contribution to sleep research analysis. The data show that specific temporal patterns can be identified within scored sleep events. A high number of temporal patterns were detected in data from patient and healthy subject’s polysomnography reports. Results indicate that the T-pattern detection algorithm discovered known sleep-patterns in manually scored apnea-patient records, sleep patterns that indicate some sort of sleep disorder. It also discovered unknown patterns that occurred exclusively in certain patient groups. Highly significant patterns were discovered in automatically scored data, based on new definitions, where certain pattern types also occurred exclusively in certain patient groups.

Paper presented at Measuring Behavior 2005 , 5^th International Conference on Methods and Techniques in Behavioral Research, 30 August - 2 September 2005, Wageningen, The Netherlands.

© 2005 Noldus Information Technology bv