Breathing rate estimation during sleep using audio signal analysis

Clinical validation of a contactless respiration rate monitor

Respiratory rate (RR) is an often underestimated and underreported vital sign with tremendous clinical value. As a predictor of cardiopulmonary arrest, chronic obstructive pulmonary disease (COPD) exacerbation or indicator of health state for example in COVID-19 patients, respiratory rate could be especially valuable in remote long-term patient monitoring, which is challenging to implement. Contactless devices for home use aim to overcome these challenges. In this study, the contactless Sleepiz One+ respiration monitor for home use during sleep was validated against the thoracic effort belt. The agreement of instantaneous breathing rate and breathing rate statistics between the Sleepiz One+ device and the thoracic effort belt was initially evaluated during a 20-min sleep window under controlled conditions (no body movement) on a cohort of 19 participants and secondly in a more natural setting (uncontrolled for body movement) during a whole night on a cohort of 139 participants. Excellent agreement was shown for instantaneous breathing rate to be within 3 breaths per minute (Brpm) compared to thoracic effort band with an accuracy of 100% and mean absolute error (MAE) of 0.39 Brpm for the setting controlled for movement, and an accuracy of 99.5% with a MAE of 0.48 Brpm for the whole night measurement, respectively. Excellent agreement was also achieved for the respiratory rate statistics over the whole night with absolute errors of 0.43, 0.39 and 0.67 Brpm for the 10th, 50th and 90th percentiles, respectively. Based on these results we conclude that the Sleepiz One+ can estimate instantaneous respiratory rate and its summary statistics at high accuracy in a clinical setting. Further studies are required to evaluate the performance in the home environment, however, it is expected that the performance is at similar level, as the measurement conditions for the Sleepiz One+ device are better at home than in a clinical setting.

Estimation of Respiratory Rate from Breathing Audio

The COVID-19 pandemic has fueled exponential growth in the adoption of remote delivery of primary, specialty, and urgent health care services. One major challenge is the lack of access to physical exam including accurate and inexpensive measurement of remote vital signs. Here we present a novel method for machine learning-based estimation of patient respiratory rate from audio. There exist non-learning methods but their accuracy is limited and work using machine learning known to us is either not directly useful or uses non-public datasets. We are aware of only one publicly available dataset which is small and which we use to evaluate our algorithm. However, to avoid the overfitting problem, we expand its effective size by proposing a new data augmentation method. Our algorithm uses the spectrogram representation and requires labels for breathing cycles, which are used to train a recurrent neural network for recognizing the cycles. Our augmentation method exploits the independence property of the most periodic frequency components of the spectrogram and permutes their order to create multiple signal representations. Our experiments show that our method almost halves the errors obtained by the existing (non-learning) methods. Clinical Relevance- We achieve a Mean Absolute Error (MAE) of 1.0 for the respiratory rate while relying only on an audio signal of a patient breathing. This signal can be collected from a smartphone such that physicians can automatically and reliably determine respiratory rate in a remote setting.

Applying ubiquitous sensing to estimate perceived exertion based on cardiorespiratory features

Reliable monitoring of one’s response to exercise intensity is imperative to effectively plan and manage training, but not always practical in impact sports settings. This study aimed to evaluate if an inexpensive mobile cardio-respiratory monitoring system can achieve similar performance to a metabolic cart in estimating rated perceived exertion. Eight adult men volunteered to perform treadmill tests under different conditions. Cardiorespiratory data were collected using a metabolic cart and an instrumented oral-cavity device, as well as their ratings of perceived exertion. Pearson correlation corrected for repeated measurements and stepwise regression analysis were used to observe the relationship between the cardiorespiratory features and the ratings of perceived exertion and determine the proportion of the variance of exertion that could be explained by the measurements. Minute ventilation was found to be the most associated variable to perceived exertion, closely followed by a novel metric called the audio minute volume, which can be collected by the oral-cavity device. A generalised linear model combining minute ventilation, audio minute volume, heart rate and respiration rate accounted for 64% of the variance in perceived exertion, whilst a model with only audio minute volume accounted for 56%. Our study indicates that minute ventilation is key to estimating perceived exertion during indoor running exercises. Audio minute volume was also observed to perform comparably to a lab-based metabolic cart in estimating perceived exertion. This research indicates that mobile techniques offer the potential for real-world data collection of an athlete’s physiological load and estimation of perceived exertion.

A Systematic Review of Sensing Technologies for Wearable Sleep Staging

Designing wearable systems for sleep detection and staging is extremely challenging due to the numerous constraints associated with sensing, usability, accuracy, and regulatory requirements. Several researchers have explored the use of signals from a subset of sensors that are used in polysomnography (PSG), whereas others have demonstrated the feasibility of using alternative sensing modalities. In this paper, a systematic review of the different sensing modalities that have been used for wearable sleep staging is presented. Based on a review of 90 papers, 13 different sensing modalities are identified. Each sensing modality is explored to identify signals that can be obtained from it, the sleep stages that can be reliably identified, the classification accuracy of systems and methods using the sensing modality, as well as the usability constraints of the sensor in a wearable system. It concludes that the two most common sensing modalities in use are those based on electroencephalography (EEG) and photoplethysmography (PPG). EEG-based systems are the most accurate, with EEG being the only sensing modality capable of identifying all the stages of sleep. PPG-based systems are much simpler to use and better suited for wearable monitoring but are unable to identify all the sleep stages.

Detection of Abnormal Respiration from Multiple-Input Respiratory Signals

In this paper, we propose a novel approach for the detection of abnormal signals from multiple respiration signals. An ultrawide-band (UWB) radar was used to acquire respiration signals that represent a distance from the chest to the radar sensor, i.e., shape variation of the chest due to breathing (inhaling or exhaling) activity provides quantitative information (distance values) about respiratory status. Distribution, shape, and variation of values across time provide information to determine respiratory status, one of the most important indicators of human health. In this paper, respiratory status was categorized into two classes, normal and abnormal. Abnormal respiration (apnea in this paper) was emulated by interrupting breathing activity because it is difficult to acquire real apnea from patients in hospital wards. This paper considered two cases, single and multiple respiration. In the first case, a single normal- or abnormal-respiration signal was used as input, and output was the classified status of respiration. In the second case, multiple respiration signals were simultaneously used as inputs, and we focused on determining the existence of abnormal signals in multiple respiration signals. In the case of multiple inputs, filters with varying cut-off frequency were applied to input signals followed by the analysis of output signals in response to the filters. To substantiate the proposed method, experiment results are provided. In this paper, classification results showed 93% of the successful rate in the case of multiple inputs, and results are promising for applications to monitoring systems of human respiration.

Clinical Applications of Mobile Health Wearable-Based Sleep Monitoring: Systematic Review

BACKGROUND

Sleep disorders are a major public health issue. Nearly 1 in 2 people experience sleep disturbances during their lifetime, with a potential harmful impact on well-being and physical and mental health.

OBJECTIVE

The aim of this study was to better understand the clinical applications of wearable-based sleep monitoring; therefore, we conducted a review of the literature, including feasibility studies and clinical trials on this topic.

METHODS

We searched PubMed, PsycINFO, ScienceDirect, the Cochrane Library, Scopus, and the Web of Science through June 2019. We created the list of keywords based on 2 domains: wearables and sleep. The primary selection criterion was the reporting of clinical trials using wearable devices for sleep recording in adults.

RESULTS

The initial search identified 645 articles; 19 articles meeting the inclusion criteria were included in the final analysis. In all, 4 categories of the selected articles appeared. Of the 19 studies in this review, 58 % (11/19) were comparison studies with the gold standard, 21% (4/19) were feasibility studies, 15% (3/19) were population comparison studies, and 5% (1/19) assessed the impact of sleep disorders in the clinic. The samples were heterogeneous in size, ranging from 1 to 15,839 patients. Our review shows that mobile-health (mHealth) wearable-based sleep monitoring is feasible. However, we identified some major limitations to the reliability of wearable-based monitoring methods compared with polysomnography.

CONCLUSIONS

This review showed that wearables provide acceptable sleep monitoring but with poor reliability. However, wearable mHealth devices appear to be promising tools for ecological monitoring.

A Contactless Respiratory Rate Estimation Method Using a Hermite Magnification Technique and Convolutional Neural Networks

The monitoring of respiratory rate is a relevant factor in medical applications and day-to-day activities. Contact sensors have been used mostly as a direct solution and they have shown their effectiveness, but with some disadvantages for example in vulnerable skins such as burns patients. For this reason, contactless monitoring systems are gaining increasing attention for respiratory detection. In this paper, we present a new non-contact strategy to estimate respiratory rate based on Eulerian motion video magnification technique using Hermite transform and a system based on a Convolutional Neural Network (CNN). The system tracks chest movements of the subject using two strategies: using a manually selected ROI and without the selection of a ROI in the image frame. The system is based on the classifications of the frames as an inhalation or exhalation using CNN. Our proposal has been tested on 10 healthy subjects in different positions. To compare performance of methods to detect respiratory rate the mean average error and a Bland and Altman analysis is used to investigate the agreement of the methods. The mean average error for the automatic strategy is 3.28 ± 3.33 % with and agreement with respect of the reference of ≈98%.

Sleep staging using nocturnal sound analysis

Sleep staging is essential for evaluating sleep and its disorders. Most sleep studies today incorporate contact sensors that may interfere with natural sleep and may bias results. Moreover, the availability of sleep studies is limited, and many people with sleep disorders remain undiagnosed. Here, we present a pioneering approach for rapid eye movement (REM), non-REM, and wake staging (macro-sleep stages, MSS) estimation based on sleep sounds analysis. Our working hypothesis is that the properties of sleep sounds, such as breathing and movement, within each MSS are different. We recorded audio signals, using non-contact microphones, of 250 patients referred to a polysomnography (PSG) study in a sleep laboratory. We trained an ensemble of one-layer, feedforward neural network classifiers fed by time-series of sleep sounds to produce real-time and offline analyses. The audio-based system was validated and produced an epoch-by-epoch (standard 30-sec segments) agreement with PSG of 87% with Cohen's kappa of 0.7. This study shows the potential of audio signal analysis as a simple, convenient, and reliable MSS estimation without contact sensors.

Screening Sleep Disordered Breathing with Noncontact Measurement in a Clinical Site

[abstFig src='/00290002/06.jpg' width='300' text='Respiratory rate from simulator and Kinect' ]<span class=”bold”>Background and purpose:</span>It has been considered that sleep-disordered breathing disorders, such as sleep apnea syndrome (SAS), cause an increase in the risk of cardiovascular disease or traffic accident risk, and thus early detection of SAS is important. It has been also important for medical workers at clinical sites to quantitatively evaluate the respiratory condition of hospitalized patients who are asleep in a simple method. A noncontact-type system was proposed to monitor the respiratory condition of sleeping patients and minimized patient-related stress such that medical workers could use the system for SAS screening and perform a preliminary check prior to definite diagnosis.<span class=”bold”>Method:</span>The system included Microsoft Kinect™ for windows® (Kinect), a tripod, and a PC. A depth sensor of Kinect was used to measure movement in the thorax motion. Data obtained from periodic waveforms were divided with the intervals of 1 min, and the number of peaks was used to obtain the respiratory rate. Additionally, a frequency analysis was performed to calculate the respiratory frequency from a frequency at which the maximum amplitude was observed. In Experiment 1), a METI-man® PatientSimulator (CAE healthcare) (simulator) was used to study the respiratory rate and frequency calculated from the Kinect data by gradually changing the designated respiratory rate. In Experiment 2), the respiratory condition of four sleeping subjects was monitored to calculate their respiratory rate and frequencies. Furthermore, a video camera was used to confirm periodic waveforms and spectrum features of body movements during sleep.<span class=”bold”>Results:</span>In Experiment 1), the results indicated that both the respiratory rate and frequency corresponded to the designated respiratory rate in each time zone. In Experiment 2), the results indicated that the respiratory rate of examines 1, 2, 3, and 4 corresponded to 12.79±2.44 times/min (average ± standard deviation), 16.46±4.33 times/min, 28.24±2.79 times/min, and 13.05±2.64 times/min, respectively. The findings also indicated that the frequency of examines 1, 2, 3, and 4 corresponded to 0.20±0.04 Hz, 0.26±0.06 Hz, 0.45±0.12 Hz, and 0.22±0.06 Hz, respectively. The periodic waveforms and amplitude spectra were enhanced with respect to body movements although regular waveform data were obtained after the body movement occurred.<span class=”bold”>Discussions:</span>The results indicated that body movement and posture temporarily affected monitoring of the system. However, the findings also revealed that it was possible to calculate the respiratory rate and frequency, and thus it was considered that the system was useful for monitoring the respiration confirm with the non-contact or SAS screening of patients in clinical site.