SEARCH

SEARCH BY CITATION

Keywords:

  • diagnosis;
  • portable monitors;
  • sleep apnoea

Summary

  1. Top of page
  2. Summary
  3. Introduction
  4. Materials and Methods
  5. Results
  6. Discussion
  7. Conclusion
  8. Conflicts of Interest
  9. References

Obstructive sleep apnoea is a highly prevalent but under-diagnosed disorder. The gold standard for diagnosis of obstructive sleep apnoea is inpatient polysomnography. This is resource intensive and inconvenient for the patient, and the development of ambulatory diagnostic modalities has been identified as a key research priority. SleepMinder (BiancaMed, NovaUCD, Ireland) is a novel, non-contact, bedside sensor, which uses radio-waves to measure respiration and movement. Previous studies have shown it to be effective in measuring sleep and respiration. We sought to assess its utility in the diagnosis of obstructive sleep apnoea. SleepMinder and polysomnographic assessment of sleep-disordered breathing were performed simultaneously on consecutive subjects recruited prospectively from our sleep clinic. We assessed the diagnostic accuracy of SleepMinder in identifying obstructive sleep apnoea, and how SleepMinder assessment of the apnoea–hypopnoea index correlated with polysomnography. Seventy-four subjects were recruited. The apnoea–hypopnoea index as measured by SleepMinder correlated strongly with polysomnographic measurement (= 0.90; ≤ 0.0001). When a diagnostic threshold of moderate–severe (apnoea–hypopnoea index ≥15 events h−1) obstructive sleep apnoea was used, SleepMinder displayed a sensitivity of 90%, a specificity of 92% and an accuracy of 91% in the diagnosis of sleep-disordered breathing. The area under the curve for the receiver operator characteristic was 0.97. SleepMinder correctly classified obstructive sleep apnoea severity in the majority of cases, with only one case different from equivalent polysomnography by more than one diagnostic class. We conclude that in an unselected clinical population undergoing investigation for suspected obstructive sleep apnoea, SleepMinder measurement of sleep-disordered breathing correlates significantly with polysomnography.


Introduction

  1. Top of page
  2. Summary
  3. Introduction
  4. Materials and Methods
  5. Results
  6. Discussion
  7. Conclusion
  8. Conflicts of Interest
  9. References

Obstructive sleep apnoea (OSA) is a highly prevalent disorder (Young et al., 2002), and is associated with impaired quality of life, increased risk of road traffic accidents, metabolic dysfunction, and increased cardiovascular morbidity and mortality (McNicholas and Bonsigore, 2007; Young et al., 2002). Effective treatment of OSA with continuous positive airway pressure therapy restores quality of life, improves cognition and at least partially reduces cardiovascular risk (Doherty et al., 2005; Patel et al., 2003). However, OSA remains significantly under-diagnosed, with access to diagnostic facilities one of the key factors limiting more widespread diagnosis (McNicholas, 2008; Tachibana et al., 2005).

The reference standard for diagnosis of OSA remains attended overnight polysomnography (PSG) performed in a sleep laboratory (American Academy of Sleep Medicine Task Force., 1999). However, PSG is resource intensive and costly, and is inconvenient and often uncomfortable for the patient undergoing investigation. Consequently, a clear research priority in the field of sleep medicine is the development of new technologies, which would permit convenient, outpatient-based and cost-effective investigation of OSA (Kuna et al., 2011).

SleepMinder (BiancaMed, Dublin, Ireland) is a non-contact bedside bio-motion sensor designed for the ambulatory investigation of sleep-disordered breathing. We have previously validated the use of SleepMinder as a means of measuring sleep/wake status (de Chazal et al., 2011). In this study, we sought to assess its utility in the diagnosis of OSA in a cohort of patients undergoing inpatient PSG for investigation of suspected sleep-disordered breathing.

Materials and Methods

  1. Top of page
  2. Summary
  3. Introduction
  4. Materials and Methods
  5. Results
  6. Discussion
  7. Conclusion
  8. Conflicts of Interest
  9. References

Subjects and data collection protocol

Subjects were recruited prospectively from patients referred to a dedicated sleep clinic for investigation of suspected sleep-disordered breathing. All patients attending the clinic were considered eligible for recruitment, with the exception of pregnant women, patients with chronic obstructive pulmonary disease requiring supplemental oxygen, and those patients with previously diagnosed OSA. Subjects were recruited between January and September 2010. Full institutional review board approval of the study was obtained, and all subjects gave written, informed consent.

Overnight PSG was performed using the Jaeger-Toennies 1000e system (Erich Jaeger GmbH, Hoechberg, Germany). Electroencephalogram (C4/A1, C3/A2), bilateral electrooculogram, submental electromyogram and electrocardiogram (modified lead V2) were recorded using surface electrodes. Respiration was measured through oronasal flow, and thoracic and abdominal movements. Oxygen saturation was measured by finger pulse oximetry. Snoring was recorded using a surface microphone, and body position was also monitored. All studies were performed in the sleep laboratory, and supervised throughout by an experienced sleep technologist. PSG data were scored by a single reader according to the recommendations of the American Academy of Sleep Medicine (AASM; Iber et al., 2007).

SleepMinder was connected directly into the PSG montage, and recorded simultaneously with other PSG signals. Following completion of the sleep study, proprietary software was used to process overnight data from SleepMinder, allowing measurement of sleep/wake status and respiration. Thereafter, the correlation between SleepMinder and PSG was assessed on a per-epoch basis, as well as a per-recording basis. Patients who were assessed as having slept for less than 2 h were removed from the study.

SleepMinder

SleepMinder was developed specifically to measure movement and respiration in sleeping human subjects. This sensor operates by transmitting very low-power radio-frequency energy (total average emitted power is 0.1 mW) at 5.8 GHz, and complies with international safety and regulatory guidelines for radio-frequency devices. It has two output signals called I and Q bio-motion signals. The SleepMinder sensor has previously been described in detail (de Chazal et al., 2011).

SleepMinder uses a specialized radio-frequency architecture that limits the sensing range to 1.5 m in front of the sensor. While SleepMinder measures all of the movements within the field of the sensor, during sleep most of the movement is related to respiratory effort of the torso, and this is the key to its utility in sleep apnoea detection. SleepMinder can detect respiration regardless of the subject's position in relation to the device, i.e. facing towards, facing away and sideways to the sensor. Furthermore, its range is configured so that when it is placed on one side of the bed it detects the breathing and movement of the person who is nearer the sensor in the event of co-sleeping bed partners.

Development of an apnoea–hypopnoea index (AHI) estimation algorithm based on bio-motion

The goal of this research was to evaluate an algorithm to automatically estimate AHI from the signal recorded by the bio-motion sensor. Fig. 1 shows a flow diagram of the classification algorithm we have developed, which uses both respiratory information and sleep/wake decisions from the sensor to estimate AHI. The apnoea detection algorithm was developed in a previous study (Zaffaroni et al., 2009).

image

Figure 1. Overall module diagram for apnoea–hypopnoea index (AHI) detection using the SleepMinder bio-motion sensor. The sleep/wake algorithm processes the sensor channels to determine periods of movement (movement flag) and 30-s epochs of sleep/wake labels (sleep/wake flags). Next, the event detection algorithm determines periods of apnoea and hypopnoea. Finally, AHI is calculated from the sleep/wake analysis and the respiratory event detections.

Download figure to PowerPoint

The processing of the bio-motion signal begins with SleepMinder's proprietary sleep/wake module. Initially, all movements in the non-contact bio-motion data are detected and mapped to activity counts. The movement flag and activity count signals are used to recognize periods of sleep and wake using movement features only. A set of features for each 30-s epoch is used in a classifier algorithm to label sleep and wake epochs. In addition, a simple presence/absence detector is employed to generate a presence mask to the sleep/wake annotations. The sleep/wake classification algorithm is described in more detail elsewhere (de Chazal et al., 2011).

In the event of gross body movement, the breathing signal will be masked by a much higher amplitude signal, and apnoea detection cannot be performed. Therefore, sections of absence, wake and gross body movement during sleep (as estimated by the sleep/wake module) are excluded from further analysis. However, smaller level movements typically associated with sleep-disordered breathing, such as gasping, are tolerated by the algorithm and are included in the apnoea detection analysis. The total remaining time will be referred to as ‘total analysis time’, representing a close approximation of the total sleep time estimated by the sleep/wake algorithm module, and is employed in the final calculation of the AHI.

Fig. 2 shows a block diagram of the sleep-disordered breathing event detection algorithm. The input signal for the apnoea detection module is represented by a series of sections of breathing signal that were previously classified as sleep. The envelope of the respiration signal is calculated by performing peak and trough detection, and by interpolating the series of the resulting peak-to-peak amplitudes, time-stamped with the time of each individual breath occurrence (defined as the midpoint between a peak and its relevant trough). An initial selection of potential apnoeic events is obtained by calculating the correlation between the envelope of the respiration signal and a set of apnoea templates. A potential event is defined as a section of the correlation feature that exceeds a threshold t_corr of 0.5 for a time t ≥ 10 s. The amplitude of the bio-motion signal is proportional to the cross-sectional area of the moving target and to the displacement of the moving target. It is highly correlated to that of the chest and abdominal bands, as well as the flow and sum signals (Shouldice et al., 2010), and can be considered an approximated measure of the breathing effort of the patient. An example of SleepMinder respiration signal versus PSG sum channel demonstrating a series of apnoeic events is displayed in Fig. 3.

image

Figure 2. Summary block diagram of the sleep-disordered breathing event detection algorithm. The envelope of the respiratory signal is determined and matched against a library of apnoea templates. Potential apnoea events are identified and then processed by the event morphological analysis module to determine detected events. Finally, the detected events are matched against epochs of sleep and the apnoea–hypopnoea index (AHI) value calculated.

Download figure to PowerPoint

image

Figure 3. An example of a comparison of the SleepMinder respiration signal and the polysomnographic (PSG) sum signal, demonstrating a series of apnoeic events in a subject with severe OSA. The amplitude of both signals has been normalized.

Download figure to PowerPoint

The morphology of the envelope of each potential apnoeic event is analysed by estimating the ratio between the residual breathing effort (the minimum of the envelope) and the baseline breathing effort (the amplitude of the envelope prior to the suspected event). In keeping with the current AASM definitions of apnoea and hypopnea, an individual event was accepted as a true detection whenever the estimated residual breathing measure dropped to below 50% of its estimated baseline value for a time longer than 10 s, with the AHI estimated from the total number of detected events divided by the total measured sleep time.

Performance measurement

The performance metrics used to assess the performance of the AHI estimator were overall correlated with expert-read PSG AHI, accuracy, sensitivity, specificity, positive predictive value, negative predictive value and Cohen's kappa coefficient in the diagnosis and staging of sleep-disordered breathing, and the area under the receiver operator curve (ROC) when compared with PSG. Positive (apnoeic) and negative subjects were defined by comparing the expert-scored AHI against clinical thresholds of 5, 15 and 30 events h−1. Bland–Altman plots were used to provide insight into the performance of classification systems between estimated and actual AHI values.

Results

  1. Top of page
  2. Summary
  3. Introduction
  4. Materials and Methods
  5. Results
  6. Discussion
  7. Conclusion
  8. Conflicts of Interest
  9. References

Patient characteristics

A total of 75 consecutive eligible subjects were enrolled during the study period between January and September 2010. One subject was excluded from the analysis on the basis of extremely short sleep duration (<45 min). Data from PSG and SleepMinder were compared for the remaining 74 patients. 80% of the study cohort was male, 55.4% were obese (body mass index >30 kg m−2), and the mean age was 49.9 years. Just under 20% had no evidence of sleep-disordered breathing, while 28.4% had mild OSA (AHI 5–15 events h−1), 25.7% moderate OSA (AHI 15–30 events h−1) and 27% severe OSA (AHI > 30 events h−1). Patient characteristics are summarized in Table 1.

Table 1. Summary of patient characteristics
Patient characteristics 
  1. Data are presented as percentage of cohort, or mean (SD).

  2. AHI, apnoea–hypopnoea index; BMI, body mass index; OSA, obstructive sleep apnoea; T2DM, type 2 diabetes mellitus.

Total (% female)74 (20.3)
Age (years)49.9 (12.3)
BMI (kg m−2)31.3 (6.2)
AHI (events h−1)26.1 (28.5)
OSA severity (%)
No OSA18.9
Mild OSA28.4
Moderate OSA25.7
Severe OSA27
Hypertension (%)20.3
T2DM (%)10.8
Statin therapy (%)17.6

AHI estimation with SleepMinder

The AHI estimation algorithm was tested against the gold standard of full PSG on the undisclosed dataset of 74 patient recordings, and yielded a correlation of 90% with < 0.001 (Fig. 4). The Bland–Altman plot (Fig. 5) showed good agreement between the SleepMinder and the PSG estimates of AHI. No significant bias was observed, and the mean difference between SleepMinder AHI estimates versus PSG was 0.05 events h−1, with a standard deviation of 12.7 events h−1. In 73% of the patient population, the SleepMinder AHI estimates did not vary by more than 10 events h−1 from PSG.

image

Figure 4. SleepMinder versus polysomnography (PSG) apnoea–hypopnoea index (AHI) estimates.

Download figure to PowerPoint

image

Figure 5. Bland–Altman plot of SleepMinder versus polysomnography (PSG) apnoea–hypopnoea index (AHI) estimates.

Download figure to PowerPoint

ROC were determined for three different PSG diagnostic thresholds of AHI ≥ 5, AHI ≥ 15 and AHI ≥ 30, and are presented in Fig. 6. Table 2 summarizes the performance of the algorithm in terms of sensitivity, specificity, positive and negative predictive values, area under the curve and Cohen's kappa for each of the above-mentioned clinical diagnostic thresholds. When a diagnostic threshold of moderate–severe OSA (i.e. AHI > 15 events h−1) was used, SleepMinder displayed a sensitivity of 90%, a specificity of 92% and an accuracy of 91% in the diagnosis of sleep-disordered breathing. Similarly, applying this threshold, the ROC yielded an area under the curve value of 0.97.

Table 2. Performance metrics at thresholds of expert AHI of 5, 15 and 30 events h−1
 AccSensSpecPPVNPVAUCk
  1. Acc, accuracy; Sens, sensitivity; Spec, specificity; PPV, positive predictivity value; NPV, negative predictivity value are expressed as a percentage value; AUC, area under the curve, and k, Cohen's kappa are expressed as decimal values; AHI, apnoea–hypopnoea index.

AHI ≥ 587984786880.900.54
AHI ≥ 1591909292890.970.81
AHI ≥ 3088848973940.960.70
image

Figure 6. The ROCs for SleepMinder versus polysomnographic (PSG) AHI estimates. Three curves are shown with the diagnostic threshold of the expert-annotated AHI set at 5, 15 and 30 events h−1, respectively. Best performance is achieved at a threshold of AHI = 15, with an area under the curve value of 0.97.

Download figure to PowerPoint

When analysing the potential for SleepMinder to classify subjects into normal (AHI < 5), mild (5 ≤ AHI < 15), moderate (15 ≤ AHI < 30) and severe (AHI ≥ 30) OSA, 49 out of 74 patients (66%) were correctly classified, and in only one case the SleepMinder diagnosis differed from equivalent PSG by more than one diagnostic class (Table 3).

Table 3. SleepMinder versus PSG output, based on a four-class severity classification
PSGSleepMinder
NormalMildModerateSevereTotal
  1. Normal breathing (AHI < 5); mild (5 ≤ AHI < 15); moderate (15 ≤ AHI < 30); and severe (AHI ≥ 30) OSA.

  2. Results are presented as number of patients (percentage).

  3. PSG, polysomnography.

Normal890017
Mild1152119
Moderate0410519
Severe0031619

Discussion

  1. Top of page
  2. Summary
  3. Introduction
  4. Materials and Methods
  5. Results
  6. Discussion
  7. Conclusion
  8. Conflicts of Interest
  9. References

The development and validation of new technologies facilitating ambulatory diagnosis is a clear research priority in sleep medicine (Kuna et al., 2011). In 1993, the estimated prevalence of moderate–severe sleep-disordered breathing within the Wisconsin sleep cohort was 9% in middle-aged men and 4% in middle-aged women (Young et al., 1993). With an evolving obesity epidemic affecting the developed world, these figures are likely to represent significant underestimates (Young et al., 2002), but even these relatively conservative estimates represent a burden of disease that significantly exceeds current conventional diagnostic capacity (Tachibana et al., 2005). Furthermore, emerging evidence suggests that high-quality home-based testing may yield similar outcomes to inpatient care in selected populations (Skomro et al., 2010).

We aimed to assess the utility of a non-contact, bio-motion sensor device in the diagnosis of sleep-disordered breathing. Use of a device such as SleepMinder is attractive for a number of reasons. First, its non-contact and unobtrusive nature potentially allows more normal sleep than traditional diagnostic modalities. Second, it requires minimal set-up, and hence has the potential to allow for convenient domiciliary testing of sleep-disordered breathing. Finally, it can facilitate study of several nights sleep with no significant additional cost. Before being applied in a clinical setting, however, these potential benefits clearly need to be matched to proven diagnostic efficacy.

Using a representative sleep clinic population undergoing inpatient investigation for possible OSA, we found SleepMinder estimation of sleep-disordered breathing correlated well with PSG measurement. This was particularly so when a diagnostic threshold of moderate–severe OSA (i.e. AHI > 15 events h−1) was used, with SleepMinder displaying a high degree of accuracy among these patients. This was further illustrated by the ROC for SleepMinder using this threshold, with an area under the curve of 0.97. A key point is that although agreement on Bland–Altman plots between SleepMinder and PSG measures of AHI diminished somewhat as OSA severity increased, importantly the device allowed the correct categorization of OSA severity in the majority of subjects, with only one patient misclassified by more than one stage. Some areas warranting further investigation were also identified; for example, AHI was significantly overestimated in one subject with periodic limb movement disorder, and underestimated in a subject with pronounced oxyhaemoglobin desaturation but relatively preserved respiratory effort. Another important limitation of the current data is the in-hospital setting of the studies; future validation studies are required to demonstrate the utility of SleepMinder in a domiciliary setting.

This study focused on a cohort attending a dedicated sleep clinic, which clearly may not be representative of the general population. However, this is in keeping with AASM clinical guidelines for the use of unattended portable monitors in the diagnosis of OSA, which recommend against the use of type 3 and type 4 devices in the investigation of asymptomatic individuals (Collop et al., 2007). These guidelines further suggest that such devices should be used particularly in subjects with a high clinical pre-test probability of significant sleep-disordered breathing. Overall, SleepMinder appears to compare favourably with other modalities used in portable monitoring, with similar performance metrics to established devices (Erman et al., 2007; Zou et al., 2006). For example, the WatchPAT-100 has been reported to have an area under the curve of 0.92 using a PSG AHI threshold of 15 events h−1 (Zou et al., 2006), while the ApneaLink device is reported to have a sensitivity of 91% and specificity of 95% using the same cut-off (Erman et al., 2007).

Conclusion

  1. Top of page
  2. Summary
  3. Introduction
  4. Materials and Methods
  5. Results
  6. Discussion
  7. Conclusion
  8. Conflicts of Interest
  9. References

SleepMinder, a non-contact bio-motion sensor, compares well with simultaneous inpatient, attended PSG in the evaluation of suspected OSA, and appears particularly effective at identifying subjects with moderate–severe sleep-disordered breathing. Future validation studies should focus on the domiciliary use of the device and the potential benefits of multi-night evaluation of sleep.

Conflicts of Interest

  1. Top of page
  2. Summary
  3. Introduction
  4. Materials and Methods
  5. Results
  6. Discussion
  7. Conclusion
  8. Conflicts of Interest
  9. References

P. de Chazal, C. Heneghan, E. O'Hare and A. Zaffaroni are employees of BiancaMed. The SleepMinder devices used in the study were provided at no cost to St Vincent's University Hospital. P. Boyle, B. Kent, G. O'Connell, M. Pallin, and W. McNicholas have no conflicts of interest to declare.

References

  1. Top of page
  2. Summary
  3. Introduction
  4. Materials and Methods
  5. Results
  6. Discussion
  7. Conclusion
  8. Conflicts of Interest
  9. References
  • American Academy of Sleep Medicine Task Force. Sleep-related breathing disorders in adults: recommendations for syndrome definition and measurement techniques in clinical research. Sleep, 1999, 22: 667689.
  • de Chazal, P., Fox, N., O'Hare, E. et al. Sleep/wake measurement using a non-contact biomotion sensor. J. Sleep Res., 2011, 20: 356366.
  • Collop, N. A., Anderson, W. M., Boehlecke, B. et al. Clinical guidelines for the use of unattended portable monitors in the diagnosis of obstructive sleep apnea in adult patients. Portable Monitoring Task Force of the American Academy of Sleep Medicine. J. Clin. Sleep Med., 2007, 3: 737747.
  • Doherty, L. S., Kiely, J. L., Swan, V. and McNicholas, W. T. Long-term effects of nasal continuous positive airway pressure therapy on cardiovascular outcomes in sleep apnea syndrome. Chest, 2005, 127: 20762084.
  • Erman, M. K., Stewart, D., Einhorn, D., Gordon, N. and Casal, E. Validation of the ApneaLink for the screening of sleep apnea: a novel and simple single-channel recording device. J. Clin. Sleep Med., 2007, 3: 387392.
  • Iber, C., Ancoli-Israel, S., Chesson, A. and Quan, S. The AASM Manual for the Scoring of Sleep and Associated Events: Rules, Terminology and Technical Specifications, 1st edn. American Academy of Sleep Medicine, Westchester: IL, 2007.
  • Kuna, S. T., Badr, M. S., Kimoff, R. J. et al. An official ATS/AASM/ACCP/ERS workshop report: research priorities in ambulatory management of adults with obstructive sleep apnea. Proc. Am. Thorac. Soc., 2011, 8: 116.
  • McNicholas, W. T. Diagnosis of obstructive sleep apnea in adults. Proc. Am. Thorac. Soc., 2008, 5: 154160.
  • McNicholas, W. T. and Bonsigore, M. R. Sleep apnoea as an independent risk factor for cardiovascular disease: current evidence, basic mechanisms and research priorities. Eur. Respir. J., 2007, 29: 156178.
  • Patel, S. R., White, D. P., Malhotra, A., Stanchina, M. L. and Ayas, N. T. Continuous positive airway pressure therapy for treating sleepiness in a diverse population with obstructive sleep apnea: results of a meta-analysis. Arch. Intern. Med., 2003, 163: 565571.
  • Shouldice, R. B., Heneghan, C., Petres, G. et al. Real time breathing rate estimation from a non contact biosensor. Conf. Proc. IEEE Eng. Med. Biol. Soc., 2010, 2010: 630633.
  • Skomro, R. P., Gjevre, J., Reid, J. et al. Outcomes of home-based diagnosis and treatment of obstructive sleep apnea. Chest, 2010, 138: 257263.
  • Tachibana, N., Ayas, N. T. and White, D. P. A quantitative assessment of sleep laboratory activity in the United States. J. Clin. Sleep Med., 2005, 1: 2326.
  • Young, T., Palta, M., Dempsey, J., Skatrud, J., Weber, S. and Badr, S. The occurrence of sleep-disordered breathing among middle-aged adults. N. Engl. J. Med., 1993, 328: 12301235.
  • Young, T., Peppard, P. E. and Gottlieb, D. J. Epidemiology of obstructive sleep apnea: a population health perspective. Am. J. Respir. Crit. Care Med., 2002, 165: 12171239.
  • Zaffaroni, A., De Chazal, P., Heneghan, C., Boyle, P., Mppm, P. R. and McNicholas, W. T. SleepMinder: an innovative contact-free device for the estimation of the apnoea-hypopnoea index. Conf. Proc. IEEE Eng. Med. Biol. Soc., 2009, 2009: 70917094.
  • Zou, D., Grote, L., Peker, Y., Lindblad, U. and Hedner, J. Validation a portable monitoring device for sleep apnea diagnosis in a population based cohort using synchronized home polysomnography. Sleep, 2006, 29: 367374.