Evaluation Of Patients With Obstructive Sleep Apnea-Hypopnea Syndrome (OSAHS) Based on Nonlinear Analysis Of Respiratory Signals

Obstructive Sleep Apnea Syndrome (OSAS)

OSAHS, nonlinear analysis, polysomnography, biosignal analysis, DFA, APEN, LLE

Objective: Obstructive Sleep Apnea-Hypopnea Syndrome (OSAHS) is a common sleep disorder requiring the time and money consuming full polysomnography to be diagnosed. Alternative methods for initial evaluation are sought. The investigators aim was the prediction of Apnea-Hypopnea Index (AHI) in patients suspected to suffer from OSAHS using two models based on nonlinear analysis of three biosignals during sleep. Methods: One hundred patients referred to a Sleep Unit underwent full polysomnography. Three nonlinear indices (Largest Lyapunov Exponent, Detrended Fluctuation Analysis and Approximate Entropy) were extracted from three biosignals (airflow from a nasal cannula, thoracic movement and Oxygen saturation) providing input to a data mining application for the creation of predictive models for AHI.

Patients referred to the Sleep Unit of a tertiary hospital in northern Greece during the years 2005-2008 and who accepted to sign the informed consent form were included in the study. One out of every five consecutive patients was selected in order to ensure randomization. The study protocol was approved by the ethics committee of the hospital. All the subjects reported symptoms consistent with OSAHS and had no significant comorbidities. The presence of dementia, neuromuscular disorders, overlap syndrome or severe cardiac problems was an exclusion criterion for the participants. The subjects underwent full overnight attended polysomnography (Somnologica 7000, Flaga; Iceland) according to standard criteria including respiratory recordings of thoracic and abdominal movements, nasal flow by pressure cannula, snoring, and arterial oxygen saturation using pulse oximetry. Apnea and hypopnea were defined in accordance with standard used criteria. All the recordings were manually scored by the same experienced medical doctor. Three nonlinear indices (Largest Lyapunov Exponent-LLE, Detrended Fluctuation Analysis-DFA and Approximate Entropy-APEN) were extracted from two respiratory signals (nasal cannula flow-F and thoracic belt movement-T). The oxygen saturation signal (SpO2) from pulse oximetry was also selected. The above signals had a mean duration of 317.5 minutes and were first exported in European Data Format (EDF) to be further processed with the use of signal processing software (Matlab by Mathworks Inc.) in personal computers. The LLE calculation required the use of a command line application by Rosenstein et al as well as a spreadsheet program (Microsoft Excel). The basic statistical analysis was performed with the use of SPSS for Windows, Version 15.0 (SPSS Inc, Chicago, Illinois). Correlations between the studied or derived parameters were explored with the Pearson's correlation test and differences in the mean observed values between the various OSAHS severity groups were analyzed using the Student's t-test. The statistical significance level was set at p\<0.05. The predictive model was created by utilizing the linear regression tool.

Greece