Estimation of macro sleep stages from whole night audio analysis.

During routine sleep diagnostic procedure, sleep is broadly divided into three states: rapid eye movement (REM), non-REM (NREM) states, and wake, frequently named macro-sleep stages (MSS). In this study, we present a pioneering attempt for MSS detection using full night audio analysis. Our working hypothesis is that there might be differences in sound properties within each MSS due to breathing efforts (or snores) and body movements in bed. In this study, audio signals of 35 patients referred to a sleep laboratory were recorded and analyzed. An additional 178 subjects were used to train a probabilistic time-series model for MSS staging across the night. The audio-based system was validated on 20 out of the 35 subjects. System accuracy for estimating (detecting) epoch-by-epoch wake/REM/NREM states for a given subject is 74% (69% for wake, 54% for REM, and 79% NREM). Mean error (absolute difference) was 36±34 min for detecting total sleep time, 17±21 min for sleep latency, 5±5% for sleep efficiency, and 7±5% for REM percentage. These encouraging results indicate that audio-based analysis can provide a simple and comfortable alternative method for ambulatory evaluation of sleep and its disorders.

Obstructive sleep apnea severity estimation: Fusion of speech-based systems.

Obstructive sleep apnea (OSA) is a common sleep-related breathing disorder. Previous studies associated OSA with anatomical abnormalities of the upper respiratory tract that may be reflected in the acoustic characteristics of speech. We tested the hypothesis that the speech signal carries essential information that can assist in early assessment of OSA severity by estimating apnea-hypopnea index (AHI). 198 men referred to routine polysomnography (PSG) were recorded shortly prior to sleep onset while reading a one-minute speech protocol. The different parts of the speech recordings, i.e., sustained vowels, short-time frames of fluent speech, and the speech recording as a whole, underwent separate analyses, using sustained vowels features, short-term features, and long-term features, respectively. Applying support vector regression and regression trees, these features were used in order to estimate AHI. The fusion of the outputs of the three subsystems resulted in a diagnostic agreement of 67.3% between the speech-estimated AHI and the PSG-determined AHI, and an absolute error rate of 10.8 events/hr. Speech signal analysis may assist in the estimation of AHI, thus allowing the development of a noninvasive tool for OSA screening.

Can we discriminate between apnea and hypopnea using audio signals?

Obstructive sleep apnea (OSA) affects up to 14% of the population. OSA is characterized by recurrent apneas and hypopneas during sleep. The apnea-hypopnea index (AHI) is frequently used as a measure of OSA severity. In the current study, we explored the acoustic characteristics of hypopnea in order to distinguish it from apnea. We hypothesize that we can find audio-based features that can discriminate between apnea, hypopnea and normal breathing events. Whole night audio recordings were performed using a non-contact microphone on 44 subjects, simultaneously with the polysomnography study (PSG). Recordings were segmented into 2015 apnea, hypopnea, and normal breath events and were divided to design and validation groups. A classification system was built using a 3-class cubic-kernelled support vector machine (SVM) classifier. Its input is a 36-dimensional audio-based feature vector that was extracted from each event. Three-class accuracy rate using the hold-out method was 84.7%. A two-class model to separate apneic events (apneas and hypopneas) from normal breath exhibited accuracy rate of 94.7%. Here we show that it is possible to detect apneas or hypopneas from whole night audio signals. This might provide more insight about a patient's level of upper airway obstruction during sleep. This approach may be used for OSA severity screening and AHI estimation.

Breathing rate estimation during sleep using audio signal analysis.

Sleep is associated with important changes in respiratory rate and ventilation. Currently, breathing rate (BR) is measured during sleep using an array of contact and wearable sensors, including airflow sensors and respiratory belts; there is need for a simplified and more comfortable approach to monitor respiration. Here, we present a new method for BR evaluation during sleep using a non-contact microphone. The basic idea behind this approach is that during sleep the upper airway becomes narrower due to muscle relaxation, which leads to louder breathing sounds that can be captured via ambient microphone. In this study we developed a signal processing algorithm that emphasizes breathing sounds, extracts breathing-related features, and estimates BR during sleep. A comparison between audio-based BR estimation and BR calculated using the traditional (gold-standard) respiratory belts during in-laboratory polysomnography (PSG) study was performed on 204 subjects. Pearson's correlation between subjects' averaged BR of the two approaches was R=0.97. Epoch-by-epoch (30 s) BR comparison revealed a mean relative error of 2.44% and Pearson's correlation of 0.68. This study shows reliable and promising results for non-contact BR estimation.

Breath-by-breath detection of apneic events for OSA severity estimation using non-contact audio recordings.

Obstructive sleep apnea (OSA) is a prevalent sleep disorder, characterized by recurrent episodes of upper airway obstructions during sleep. We hypothesize that breath-by-breath audio analysis of the respiratory cycle (i.e., inspiration and expiration phases) during sleep can reliably estimate the apnea hypopnea index (AHI), a measure of OSA severity. The AHI is calculated as the average number of apnea (A)/hypopnea (H) events per hour of sleep. Audio signals recordings of 186 adults referred to OSA diagnosis were acquired in-laboratory and at-home conditions during polysomnography and WatchPat study, respectively. A/H events were automatically segmented and classified using a binary random forest classifier. Total accuracy rate of 86.3% and an agreement of κ=42.98% were achieved in A/H event detection. Correlation of r=0.87 (r=0.74), diagnostic agreement of 76% (81.7%), and average absolute difference AHI error of 7.4 (7.8) (events/hour) were achieved in in-laboratory (at-home) conditions, respectively. Here we provide evidence that A/H events can be reliably detected at their exact time locations during sleep using non-contact audio approach. This study highlights the potential of this approach to reliably evaluate AHI in at home conditions.

Detection of breathing sounds during sleep using non-contact audio recordings.

Evaluation of respiratory activity during sleep is essential in order to reliably diagnose sleep disorder breathing (SDB); a condition associated with serious cardio-vascular morbidity and mortality. In the current study, we developed and validated a robust automatic breathing-sounds (i.e. inspiratory and expiratory sounds) detection system of audio signals acquired during sleep. Random forest classifier was trained and tested using inspiratory/expiratory/noise events (episodes), acquired from 84 subjects consecutively and prospectively referred to SDB diagnosis in sleep laboratory and in at-home environment. More than 560,000 events were analyzed, including a variety of recording devices and different environments. The system's overall accuracy rate is 88.8%, with accuracy rate of 91.2% and 83.6% in in-laboratory and at-home environments respectively, when classifying between inspiratory, expiratory, and noise classes. Here, we provide evidence that breathing-sounds can be reliably detected using non-contact audio technology in at-home environment. The proposed approach may improve our understanding of respiratory activity during sleep. This in return, will improve early SDB diagnosis and treatment.

Detection of Obstructive sleep apnea in awake subjects by exploiting body posture effects on the speech signal.

Obstructive sleep apnea (OSA) is a common sleep disorder. OSA is associated with several anatomical and functional abnormalities of the upper airway. It was shown that these abnormalities in the upper airway are also likely to be the reason for increased rate of apneic events in the supine position. Functional and structural changes in the vocal tract can affect the acoustic properties of speech. We hypothesize that acoustic properties of speech that are affected by body position may aid in distinguishing between OSA and non-OSA patients. We aimed to explore the possibility to differentiate OSA and non-OSA patients by analyzing the acoustic properties of their speech signal in upright sitting and supine positions. 35 awake patients were recorded while pronouncing sustained vowels in the upright sitting and supine positions. Using linear discriminant analysis (LDA) classifier, accuracy of 84.6%, sensitivity of 92.7%, and specificity of 80.0% were achieved. This study provides the proof of concept that it is possible to screen for OSA by analyzing and comparing speech properties acquired in upright sitting vs. supine positions. An acoustic-based screening system during wakefulness may address the growing needs for a reliable OSA screening tool; further studies are needed to support these findings.

OSA severity assessment based on sleep breathing analysis using ambient microphone.

In this paper, an audio-based system for severity estimation of obstructive sleep apnea (OSA) is proposed. The system estimates the apnea-hypopnea index (AHI), which is the average number of apneic events per hour of sleep. This system is based on a Gaussian mixture regression algorithm that was trained and validated on full-night audio recordings. Feature selection process using a genetic algorithm was applied to select the best features extracted from time and spectra domains. A total of 155 subjects, referred to in-laboratory polysomnography (PSG) study, were recruited. Using the PSG's AHI score as a gold-standard, the performances of the proposed system were evaluated using a Pearson correlation, AHI error, and diagnostic agreement methods. Correlation of R=0.89, AHI error of 7.35 events/hr, and diagnostic agreement of 77.3% were achieved, showing encouraging performances and a reliable non-contact alternative method for OSA severity estimation.

Sleep-quality assessment from full night audio recordings of sleep apnea patients.

In this work, a novel system (method) for sleep quality analysis is proposed. Its purpose is to assist an alternative non-contact method for detecting and diagnosing sleep related disorders based on acoustic signal processing. In this work, audio signals of 145 patients with obstructive sleep apnea were recorded (more than 1000 hours) in a sleep laboratory and analyzed. The method is based on the assumption that during sleep the respiratory efforts are more periodically patterned and consistent relative to a waking state; furthermore, the sound intensity of those efforts is higher, making the pattern more noticeable relative to the background noise level. The system was trained on 50 subjects and validated on 95 subjects. The system accuracy for detecting sleep/wake state is 82.1% (epoch by epoch), resulting in 3.9% error (difference) in detecting sleep latency, 11.4% error in estimating total sleep time, and 11.4% error in estimating sleep efficiency.

Adenotonsillectomy improves slow-wave activity in children with obstructive sleep apnoea.

The aim of the present study was to estimate slow-wave activity (SWA), a marker of sleep homeostasis, in children with obstructive sleep apnoea (OSA) before and after adenotonsillectomy (AT) compared with untreated OSA children (comparison group). 14 children with OSA (mean ± sd age 6.4 ± 2.5 yrs; apnoea-hypopnoea index (AHI) 10.0 ± 10.3 events·h⁻¹) who underwent AT were consecutively recruited to the study. The comparison group comprised six retrospectively recruited children (age 5.4 ± 2.2 yrs; AHI 9.4 ± 7.6 events·h⁻¹) with OSA that did not undergo treatment. Electroencephalogram (derivation C3/A2) was analysed using spectral and waveform analysis to determine SWA energy and slow-wave slope. The same procedure was repeated 5.4 and 19 months later for the AT and comparison groups, respectively. AT improved respiration without a change in duration of sleep stages. Following AT, >50% elevation of SWA during the first two sleep cycles (p<0.01) and a more physiological decay of SWA across the night (p<0.0001) were noted. The slow-wave slope increased by >30% following AT (p<0.03). No significant changes were found in SWA in the comparison group. Sleep homeostasis is considerably impaired in pre-pubescent children with OSA. AT restores more physiological sleep homeostasis in children with OSA. SWA analysis may provide a useful addition to standard sleep-stage analyses in children with OSA.

ECG signal compression using analysis by synthesis coding.

In this paper, an elecrocardiogram (ECG) compression algorithm, called analysis by synthesis ECG compressor (ASEC), is introduced. The ASEC algorithm is based on analysis by synthesis coding, and consists of a beat codebook, long and short-term predictors, and an adaptive residual quantizer. The compression algorithm uses a defined distortion measure in order to efficiently encode every heartbeat, with minimum bit rate, while maintaining a predetermined distortion level. The compression algorithm was implemented and tested with both the percentage rms difference (PRD) measure and the recently introduced weighted diagnostic distortion (WDD) measure. The compression algorithm has been evaluated with the MIT-BIH Arrhythmia Database. A mean compression rate of approximately 100 bits/s (compression ratio of about 30:1) has been achieved with a good reconstructed signal quality (WDD below 4% and PRD below 8%). The ASEC was compared with several well-known ECG compression algorithms and was found to be superior at all tested bit rates. A mean opinion score (MOS) test was also applied. The testers were three independent expert cardiologists. As in the quantitative test, the proposed compression algorithm was found to be superior to the other tested compression algorithms.

The weighted diagnostic distortion (WDD) measure for ECG signal compression.

In this paper, a new distortion measure for electrocardiogram (ECG) signal compression, called weighted diagnostic distortion (WDD) is introduced. The WDD measure is designed for comparing the distortion between original ECG signal and reconstructed ECG signal (after compression). The WDD is based on PQRST complex diagnostic features (such as P wave duration, QT interval, T shape, ST elevation) of the original ECG signal and the reconstructed one. Unlike other conventional distortion measures [e.g. percentage root mean square (rms) difference, or PRD], the WDD contains direct diagnostic information and thus is more meaningful and useful. Four compression algorithms were implemented (AZTEC, SAPA2, LTP, ASEC) in order to evaluate the WDD. A mean opinion score (MOS) test was applied to test the quality of the reconstructed signals and to compare the quality measure (MOSerror) with the proposed WDD measure and the popular PRD measure. The evaluators in the MOS test were three independent expert cardiologists, who studied the reconstructed ECG signals in a blind and a semiblind tests. The correlation between the proposed WDD measure and the MOS test measure (MOSerror) was found superior to the correlation between the popular PRD measure and the MOSerror.