Subband decomposition soft-decision algorithm for heart rate variability analysis in patients with obstructive sleep apnea and normal controls

Abstract

A new method for screening of obstructive sleep apnea (OSA) is investigated. This method is based on the estimation of the energy distribution of the R–R interval (RRI) signals in the time domain. The novelty of the technique arises from the implementation of the soft-decision algorithm of subband decomposition. This soft-decision algorithm will help in finding the ratio of energy (power spectral density (PSD)) in the different frequency bands of the RRI spectrum without implementing any transform technique. Two different ratios—low-frequency/very low-frequency (LF/VLF) and low-frequency/high-frequency (LF/HF)—are used for screening normal and apnea cases. The algorithm can be implemented directly on the (RRI) raw-data or after some pre-processing and filtering steps. The training data used in this study are drawn from the MIT-trial database, while the test data are drawn from the MIT-challenge (chal) database as well as from the sleep disorders laboratory of Sultan Qaboos University (SQU) hospital. Threshold values to identify normal and OSA cases are selected using the receiver operating characteristics (ROC) on the training data. These threshold values are then used for the screening of the test data. The best classification accuracy obtained with the test data (MIT-chal and SQU data) approaches 93% using the LF/VLF ratio. In this case, the sensitivities obtained with MIT-chal and SQU data are 95% and 100%, respectively, while the specificities are 90% and 86% for the same two groups of data.