An intelligent method for identifying cardiac cycles from tracheal sounds during sleep

Abstract

The prevalence of sleep disorders is increasing and they are becoming more and more complex public health problems. Auscultation of the heart is a common procedure of medical practitioners, though it is highly subjective and depends on the expertise of the doctor. In this paper, we present a new method for identifying the cardiac cycles from the tracheal sound signal during sleep. Polysomnographic recordings of 4 female and 8 male subjects were analyzed. The tracheal sound signal was measured with a sampling rate of 11025 Hz. As the first step in the present development, the tracheal sound signal was low-pass filtered with a cut-off frequency of 50 Hz. Then, the Hilbert transform was performed to obtain the envelope, which was further smoothed with moving average filtering. Additionally, a local maximum signal was extracted from the smoothed envelope and cardiac cycles were detected at the time instances when the maximum signal value equaled the envelope value. Beat-to-beat intervals of the cardiac cycle were determined as the time between two consecutive cardiac cycle detections. The beat-to-beat intervals obtained with the developed method and from the reference method (electrocardiogram, R-R interval) were determined and collected to separate pools. Median values and 25% and 75% percentiles of beat-tobeat durations were extracted for evaluation. The method presented here provided high concordance with the reference method in all subjects. The developed method seems to be a promising tool for identifying cardiac cycles from tracheal sounds. Thus, analysis of tracheal sounds can be utilised in monitoring of the cardiovascular system with methods such as presented here. Analysis of tracheal sounds offers an interesting modality for evaluating the cardiovascular system during sleep. © 2008 Springer-Verlag.