Performance evaluation of the spectral autocorrelation function and autoregressive models for automated sleep apnea detection using single-lead ECG signal. (October 2020)

• Record Type: Journal Article
• Authors: Zarei, Asghar; Mohammadzadeh Asl, Babak
• Is Part Of: Computer methods and programs in biomedicine. Volume 195(2020)
• Publication Date: 2020-10
• Subjects: Spectral autocorrelation function -- AR Model -- Obstructive sleep apnea -- Electrocardiogram (ECG); Medicine -- Computer programs -- Periodicals; Biology -- Computer programs -- Periodicals; Computers -- Periodicals; Medicine -- Periodicals; Médecine -- Logiciels -- Périodiques; Biologie -- Logiciels -- Périodiques; Biology -- Computer programs; Medicine -- Computer programs; Periodicals; Electronic journals; 610.28
• DOI: 10.1016/j.cmpb.2020.105626
• Languages: English
• ISSNs: 0169-2607

Abstract

Highlights: A single-lead ECG based automated sleep apnea detection algorithm is proposed. AR model and spectral autocorrelation function were employed in OSA detection. Our algorithm achieved an accuracy of 93.90% in per-segment classification. Our method is superior to other state-of-the-art ECG-based OSA detection algorithms. This method can be used in the e-health applications and portable systems. Abstract: Background and objective: This paper addresses the automated recognition of obstructive sleep apnea (OSA) from the analysis of single-lead ECG signals. This is one of the most important problems that is, critical to the realization of monitoring patients with sleep apnea. Methods: In the present study, a novel solution based on autoregressive (AR) modeling of the single-lead ECG, and spectral autocorrelation function as an ECG feature extraction method is presented. The more effective features are opted by sequential forward feature selection (SFFS) technique and fed into the random forest for binary classification between the apnea and normal events. Results: Experimental results on Apnea-ECG database proved that the introduced algorithm resulted in an accuracy of 93.90% (sensitivity of 92.26% and specificity of 94.92%) in per-segment classification, which outperforms the other cutting-edge automatic OSA recognition techniques. Moreover, the proposed algorithm provided an accuracy of 97.14% (sensitivity of 95.65% and specificity of 100%) in discrimination of apnea patients from the normal subjects, which is comparable to the traditional and existing approaches. Conclusions: This study suggests that automatic OSA recognition from single-lead ECG signals is possible, which can be used as an inexpensive and low complexity burden alternative to more conventional methods such as Polysomnography.