Classification of sleep and wake using a novel minimal-contact single-channel device

Investigation of respiratory rate in patients with cystic fibrosis using a minimal-impact biomotion system

Background

In this study we tested the hypothesis that in patients with cystic fibrosis (pwCF) respiratory rate (RR) is associated with antibiotic treatment, exacerbation status, forced expiratory volume in one second (FEV1) and C-reactive protein (CRP).

Methods

Between June 2018 and May 2019, we consecutively enrolled pwCF who were referred to our hospital. We determined RR and heart rate (HR) by using the minimal-impact system VitaLog during the hospital stay. Furthermore, we performed spirometry and evaluated CRP.

Results

We included 47 patients: 20 with pulmonary exacerbation and 27 without. RR decreased in patients with exacerbation (27.5/min (6.0/min) vs. 24.4/min (6.0/min), p = 0.004) and in patients with non-exacerbation (22.5/min (5.0/min) vs. 20.9/min (3.5/min), p = 0.024). Patients with exacerbation showed higher RR than patients with non-exacerbation both at the beginning (p = 0.004) and at the end of their hospital stay (p = 0.023). During the hospital stay, HR did not change in the total cohort (66.8/min (11.0/min) vs. 66.6/min (12.0/min), p = 0.440). Furthermore, we did not find significant differences between patients with exacerbation and patients with non-exacerbation (67.0/min (12.5/min) vs. 66.5/min (10.8/min), p = 0.658). We observed a correlation of ρ = -0.36 between RR and FEV1. Moreover, we found a correlation of ρ = 0.52 between RR and CRP.

Conclusion

In pwCF requiring intravenous therapy, respiratory rate is higher at their hospital admittance and decreased by the time of discharge; it is also associated with C-reactive protein. Monitoring RR could provide important information about the overall clinical conditions of pwCF.

A deep transfer learning approach for wearable sleep stage classification with photoplethysmography

Unobtrusive home sleep monitoring using wrist-worn wearable photoplethysmography (PPG) could open the way for better sleep disorder screening and health monitoring. However, PPG is rarely included in large sleep studies with gold-standard sleep annotation from polysomnography. Therefore, training data-intensive state-of-the-art deep neural networks is challenging. In this work a deep recurrent neural network is first trained using a large sleep data set with electrocardiogram (ECG) data (292 participants, 584 recordings) to perform 4-class sleep stage classification (wake, rapid-eye-movement, N1/N2, and N3). A small part of its weights is adapted to a smaller, newer PPG data set (60 healthy participants, 101 recordings) through three variations of transfer learning. Best results (Cohen's kappa of 0.65 ± 0.11, accuracy of 76.36 ± 7.57%) were achieved with the domain and decision combined transfer learning strategy, significantly outperforming the PPG-trained and ECG-trained baselines. This performance for PPG-based 4-class sleep stage classification is unprecedented in literature, bringing home sleep stage monitoring closer to clinical use. The work demonstrates the merit of transfer learning in developing reliable methods for new sensor technologies by reusing similar, older non-wearable data sets. Further study should evaluate our approach in patients with sleep disorders such as insomnia and sleep apnoea.

Sleep assessment in cystic fibrosis patients using a minimal-impact biomotion system

PURPOSE

In our study we aimed to analyze sleep variability and activity in patients with cystic fibrosis (CF) during their hospital stay.

METHODS

Forty-three CF patients were recruited and have been divided into two subgroups: exacerbated (n = 18) and non-exacerbated (n = 25). During the course of their hospital stay we used VitaLog, a minimal-impact biomotion device, in order to determine total sleep time (TST), time in bed (TIB), sleep efficiency (SE) and intra patient standard deviation (IPSD) of TST.

RESULTS

TST was 5.1 h ± 1.5 h and ranged from 0.6h to 7.9 h.TIB was 17.7 h ± 3.8 h and ranged from 5.6h to 23.9 h. SE was 70.0% ± 17.0% and ranged from 13.6% to 98.5%. TST was higher in non-exacerbated patients (5.3 h ± 1.4 h vs. 4.8 h ± 1.6 h, p = 0.008) whereas TIB was lower in non-exacerbated patients (17.0 h ± 3.7 h vs. 18.5 h ± 3.8 h, p = 0.002). We also found that SE was better in non-exacerbated patients (73.1% ± 14.6% vs. 66.6% ± 18.8%, p = 0.002). Furthermore, we observed that IPSD of TST was higher in exacerbated patients (1.3 h ± 0.5 h vs. 0.9 h ± 0.4 h, p = 0.004).

CONCLUSION

In general, in CF patients TST was short and SE poor during the night. Furthermore, in the course of their hospital stay patients showed low activity. In exacerbated patients sleep quality was lower compared to non-exacerbated patients.

Alternative algorithms and devices in sleep apnoea diagnosis: what we know and what we expect

PURPOSE OF REVIEW

Diagnosis of sleep apnoea was performed in sleep laboratories with polysomnography. This requires a room with supervision and presence of technologists and trained sleep experts. Today, clinical guidelines in most countries recommend home sleep apnoea testing with simple systems using six signals only. If criteria for signal quality, recording conditions, and patient selection are considered, then this is a reliable test with high accuracy.

RECENT FINDINGS

Recently diagnostic tools for sleep apnoea diagnosis become even more simple: smartwatches and wearables with smart apps claim to diagnose sleep apnoea when these devices are tracking sleep and sleep quality as part of new consumer health checking. Alternative and new devices range from excellent diagnostic tools with high accuracy and full validation studies down to very low-quality tools which only result in random diagnostic reports. Due to the high prevalence of sleep apnoea, even a random diagnosis may match a real disorder sometimes.

SUMMARY

Until now, there are no metrics established how to evaluate these alternative algorithms and simple devices. Proposals for evaluating smartwatches, smartphones, single-use sensors, and new algorithms are presented. New assessments may help to overcome current limitations in sleep apnoea severity metrics. VIDEO ABSTRACT: http://links.lww.com/COPM/A28.

A novel minimal-contact biomotion method for long-term respiratory rate monitoring

PURPOSE

In this study, we assessed the diagnostic accuracy of the device VitaLog (SWG Sportwerk GmbH & Co. KG, Dortmund, Germany) for estimation of respiratory rate (RR) variability.

METHODS

VitaLog is a minimal-contact biomotion device that is placed under the mattress topper. It senses respiratory effort and body movement using a piezoelectric sensor. Diagnostic accuracy was determined in 103 patients referred to our sleep laboratory for suspected sleep-disordered breathing (SDB). SDB was defined by AHI ≥ 15/h. Results provided by VitaLog were compared with nasal flow measurement obtained by polysomnography (PSG).

RESULTS

Diagnostic accuracy of VitaLog was excellent. We obtained a correlation of r = 0.99 and a bias of 0.2 cycles per minute (cpm) between VitaLog and PSG-provided nasal flow. Detection RR variability worked nearly identically in patients with and without SDB.

CONCLUSION

VitaLog is an appropriate method for determination of RR variability based on a minimal-contact biomotion sensor. This device is easy to handle, available at low cost, and suitable for long-term monitoring in the hospital or at home.

Sleep stage classification from heart-rate variability using long short-term memory neural networks

Automated sleep stage classification using heart rate variability (HRV) may provide an ergonomic and low-cost alternative to gold standard polysomnography, creating possibilities for unobtrusive home-based sleep monitoring. Current methods however are limited in their ability to take into account long-term sleep architectural patterns. A long short-term memory (LSTM) network is proposed as a solution to model long-term cardiac sleep architecture information and validated on a comprehensive data set (292 participants, 584 nights, 541.214 annotated 30 s sleep segments) comprising a wide range of ages and pathological profiles, annotated according to the Rechtschaffen and Kales (R&K) annotation standard. It is shown that the model outperforms state-of-the-art approaches which were often limited to non-temporal or short-term recurrent classifiers. The model achieves a Cohen's k of 0.61 ± 0.15 and accuracy of 77.00 ± 8.90% across the entire database. Further analysis revealed that the performance for individuals aged 50 years and older may decline. These results demonstrate the merit of deep temporal modelling using a diverse data set and advance the state-of-the-art for HRV-based sleep stage classification. Further research is warranted into individuals over the age of 50 as performance tends to worsen in this sub-population.