//Logo Image
Author: Chih-Ming Cheng (2007-07-); recommended: Yeh-Liang Hsu (2007-07-).
Note: This article is Chapter 3 of Chih-Ming Cheng’s PhD thesis Development of a portable system for tele-monitoring of sleep in a home environment.

Chapter 3. Development of a portable device for tele-monitoring of snoring and OSAS symptoms

This chapter describes the development of a “Snoring and OSAS symptom Detector (SOD)”, which is designed for long-term monitoring at home based on the PTMS structure. The SOD is not intended to be a diagnosis device for OSAS, instead, this portable device is to be used as a home appliance as a precautionary measure for monitoring snoring and OSAS. The SOD on-line monitors snoring and intermittent snoring pattern, which is the characteristic symptom of OSAS, and stores the monitoring data. Users and caregivers can access the SOD for historical monitoring data via the Internet.

This chapter is organized as follows. Section 3.1 describes the structure of the SOD. Section 3.2 and Section 3.3 presents the algorithm design and network framework of the SOD. Section 3.4 discusses the validation results of the SOD. In summary, Section 3.5 describes the applications and benefit of the SOD.

3.1     Structure of the SOD

Figure 3-1 shows the structure of the SOD developed in this research. Similar to the PTMS structure in chapter 1, the core component of the SOD is a DDS that analyses sound signals captured by a microphone and selects snoring and OSAS symptoms using the snore and OSAS detecting algorithms. The DDS consists of a PIC server mounted on a peripheral application board. The PIC server integrates a PIC microcontroller (PIC18F6722, Microchip), EEPROM (24LC1025, Microchip) and a networking IC (RTL8019AS, Realtek). It provides networking capability and can be used as a web server.

The counts of snoring and intermittent snoring pattern per minute, as well as the duration of intermittent snoring pattern, are recorded in the Multi-Media-Card (MMC) of the DDS. Authorized remote users can request data from the DDS using an Internet web browser (through an application server) or a VB program (direct access to the DDS).

This portable tele-monitoring device provides a convenient approach to better understand and recognize sleep-related breathing disorders. Patients whose snoring patterns are classified as possible OSAS symptoms by the SOD should consult doctors for further diagnosis. For this purpose, a sound recorder is activated to record a one-minute episode for further diagnosis, once an OSAS symptom is detected. In the mean time, a low-frequency electrical stimulation can be given to the patient in order to stop the development of an apnoeic process. If an urgent situation is detected, an emergency message (mobile phone short messages or emails) can be sent to specified caregivers.

Figure 3-1. Structure of the snoring and OSAS symptoms detector

3.2     Design of the snoring and OSAS symptom detecting algorithms

As shown in Figure 3-1, an omnidirectional electret condenser type microphone is used in the SOD to record the snoring sounds. A fifth-order, low-pass filter removed high-frequency sounds. The frequency response of the microphone is within 50-30,000Hz, and the cut-off frequency of the low-pass filter is 200Hz. The analog signals are digitized at a sampling frequency of 2KHz with a 10-bit A/D converter in a PIC server. The signals are processed by a series of smoothing procedures and analyzed with the snoring and OSAS detecting algorithms. The following sections describe this procedure in details.

3.2.1        Smoothing and simplifying procedures

Figure 3-2(a) shows the input snoring sound signals. A series of smoothing and simplifying procedures further process these signals:

(1)   The differences between the input voltages and the baseline voltage are calculated (Figure 3-2(a)).

(2)   The absolute values of the differences are taken to get all positive values (Figure 3-2(b)).

(3)   A moving average filter with a window size of 20 smoothes the profile of the input voltage values.

(4)   To save computational resource in the following snoring detecting algorithm, the signals are further simplified to 10 data points per second (Figure 3-2(c)).

Figure 3-2(a). Input voltages and the baseline voltage

Figure 3-2(b). Absolute value of the differences

Figure 3-2(c). Smoothed and simplified values of the input voltage

3.2.2        Snoring detecting algorithm

After the smoothing and simplifying procedure, the snoring detecting algorithm analyzes the data for detecting and locating snores by the following steps:

(1)   Calculate the slope values of snoring signals in Figure 3-2(c), as shown in Figure 3-3(a).

(2)   Symbolize a slope value with a snoring code “1” if it is above the threshold limit and a “0” if the slope value is under the threshold limit. Figure 3-3(b) shows this “slope mapping” of the first two snores in Figure 3-3(a), using a threshold limit 0.003.

(3)   Identify snores if the slope mapping result matches the length definition of a snore (Figure 3-3(c)).

Figure 3-3(a). Slope values of snoring signals

Figure 3-3(b). Slope code of snoring sounds (Threshold limit value=0.003)

Figure 3-3(c). Snoring episodes located with the slope mapping method

To determine the length definition of a snore, 5024 snoring samples from 5 patients who were diagnosed as having sleep-related breathing disorders were collected and analyzed. Using the same threshold limit 0.003, it was found that the duration (the length of a series of snoring code “1”) of 99% of the snores ranged from 0.6 seconds to 1.8 seconds. This duration is used in the snoring detecting algorithm in SOD.

3.2.3        OSAS symptom detecting algorithm

Figure 3-4 shows a typical waveform of an OSAS symptom, where the patient stopped breathing after a series of snores and started breathing again after 34 seconds. The “Apnea-Hypopnea Index (AHI)” is an index of severity that combines apneas and hypopneas. AHI is calculated by dividing the number of apneas and hypopneas by the number of hours of sleep. In MESAM IV described in Section 3-1, intermittent snoring intervals lasting between 10-60 seconds are assumed as an intermittent snore [Penzel, et al., 1990]. An intermittent snoring interval longer than 60 seconds may be a sign that the patient stops snoring and breathes normally, and will not be assumed as an OSAS symptom. The “Intermittent Snoring Index (ISI)”, which was defined as the number of pauses in snoring lasting 10 to 60 seconds per hour, was one of the indices used in MESAM IV for automatic scoring of OSAS.

Figure 3-4. The waveform of an OSAS segment

Both AHI and ISI are calculated off-line. To identify the OSAS symptom in real time, the “Intermittent Snoring Ratio (ISR)” is defined for describing the intermittent snoring pattern for each minute. The SOD tallies intermittent snore counts (intermittent snoring intervals lasting between 10-60 seconds) and total snore counts (normal snore counts plus intermittent snore counts) within the current minute. The location of normal snore and intermittent snore is decided by the end point of the snoring episode. At the end of the minute, ISR is calculated by

        ISR = intermittent snore counts / total snore counts                            (3-1)

To consider the tendency of ISRs for the past 5 minutes (including the current minute), the “Weighted-ISR (W-ISR)” is calculated by

   (3-2)

High W-ISR represents a frequent intermittent snoring pattern in the past 5 minutes. Therefore the intermittent snore in the current minute is more likely to be OSAS symptom, instead of a sign that the patient stops snoring and breathes normally. The maximum W-ISR is 1, which means that all snores in the past 5 minutes are intermittent snores. In this research, the SOD calculates W-ISR values and classifies the intermittent snore in the past minute as OSAS symptom if the W-ISR value is higher than 0.25.

Figure 3-5 and Figure 3-6 show 6-hour records of the W-ISRs of a normal snorer and a patient who was diagnosed to have OSAS. W-ISRs are high for the OSAS patient and relatively lower for the normal snorer. Figure 3-7 shows a one-hour record of W-ISRs (4:00-5:00AM). The periods in which OSAS symptoms occur identified by the SOD using W-ISRs match well with those diagnosed by PSG using SaO2 saturation, nasal airflow and thoracic effort sensors.

Figure 3-5. W-ISRs of a normal snorer from 0:00 to 6:00AM

 Figure 3-6. W-ISRs of an OSAS patient from 0:00 to 6:00AM

Figure 3-7. The periods in which OSAS symptoms occur identified via SOD and PSG

3.3     User Interface and Network Framework

Figure 3-8 shows a prototype of the SOD developed in this research. The SOD can be easily installed by connecting the power line and the Internet cable. During monitoring, the numbers of snore counts per minute, intermittent snore counts per minute, as well as W-ISR are shown on the LCD display and stored in the MMC. Two external modules, a sound recorder and a transcutaneous electrical nerve stimulator (TENS SW32012, SHINMED), can also be connected to the SOD (Figure 3-9). Once OSAS symptoms are detected, low-frequency transcutaneous electrical nerve stimulation can be given to the patient in order to stop the development of an apnoeic process. The sound recorder is also activated to record a one-minute episode for further diagnosis.

Figure 3-8. A prototype of the SOD

Figure 3-9. External modules of SOD

Figure 3-10 shows the network framework for the SOD. Doctors and caregivers can access real-time and historical data via a remote VB program. Figure 3-11(a) shows the VB interface window designed for the SOD. After typing an IP address and clicking the “real-time” button, real-time information of snore count, intermittent snore count and W-ISR are displayed on the left side of the interface window. OSAS alarm will be flagged when W-ISR is higher than 0.25. Doctors and caregivers can monitor 3 patients on the same window. Selecting one of the IP addresses on the left side, the user can also download historical data from the SOD at this IP address (select a date and time interval on the right side of the interface window and click the “download” button). Diagnosis reports for total snore count, total intermittent snore count and total OSAS segments (in minutes) within the selected time interval can be tallied automatically. Charts of snore counts per minute versus time, intermittent snore counts per minute versus time, and OSAS segments versus time can also be drawn for further analysis (Figure 3-11(b)-(d)).

Figure 3-10. A framework for the SOD

Figure 3-11(a). VB interface window for the SOD

Figure 3-11(b). Snore counts per minute versus time

Figure 3-11(c). Intermittent snore count per minute versus time

Figure 3-11(d). OSAS segments versus time

3.4     Validation tests of the snoring and OSAS symptom detector

Two validation tests were designed to evaluate the performance of the snoring and OSAS symptom detector. The first test is for snore detecting validation, in which the SOD was used to detect snores of 5 normal snorers and 5 OSAS patients in whole-night’s sleep. All automatically detected snores by the SOD were checked by manual annotations.

One problem with manual annotation was that, the definition of snoring varies between observers. Hoffstein et al. [1996] pointed out that, the difficulty in measuring and quantifying snoring using objective criteria is that snoring is the first and foremost of a subjective perception by a listener. They studied 25 cases, all had full nocturnal polysomnography including the measurements of snoring, to compare the objective snore count from the polysomnogram and the subjective snoring count by two listeners during a 20-minute segment. In 7 out of 25 patients, the difference in subjective snore counts perceived by both listeners was larger than 25%.

To reduce the uncertainty of subjective judgments, a sound level meter (TES 1350A, TES Electrical Electronic Corp.) was used to record the sound level of each snore. A snoring segment greater than 60db was classified as a snore and segments below 60db were classified by the observers’ own judgments. A camcorder was used to record the colored LED lights of the SOD which flash when snoring episodes were detected, and the display panels of both the sound level meter and the clock. Figure 3-12 shows the test environment and test equipment.

A trained observer viewed the entire night’s sleep videos and checked each snore identified by the SOD by manual annotations. The comparison was coded as either: True Positive (TP), False Negative (FN) or False Positive (FP). Table 3-1 shows the validation results of snoring detection. The SOD showed a good performance in detecting snore and achieved an average sensitivity of 94.0% (ranged from 83.4%-99.8%) and an average positive predictive value (PPV) of 94.0% (ranged from 83.3%-99.9%).

In the second test, OSAS symptoms judged by the SOD with W-ISR were compared with the diagnosis reports by a PSG. A one-minute segment was coded TP if the SOD and PSG both detected OSAS symptoms within the minute. Similarly, a segment was coded FP or FN if only the SOD or only the PSG detected OSAS symptoms within the minute. Table 3-2 shows the validation results of locating OSAS segment of the 5 OSAS patients. The average sensitivity of locating OSAS segments was 79.6% (ranged from 62.4%-94.6%) and the average PPV was 76.3% (ranged from 39.8%-94.4%) in the OSAS patient group.

In the mean time, 360-minutes of recorded data were collected from 5 normal snorers for evaluating the success rate of recognizing normal segments (non-OSAS segments). Segments with W-ISRs below 0.25 were classified as normal segments. The average sensitivity of recognizing normal segments was 88.3% (Table 2-3), that is, 11.7% of the non-OSAS segments were falsely identified as OSAS segments.

Figure 3-12. Test environment and test equipments

Table 3-1. System performance validation results of snoring detection

Group

Patient

Total Snores

TP

FN

FP

Sensitivity

PPV

Normal snorers

Snorer A

547

456

91

73

83.4%

86.2%

Snorer B

896

778

118

24

86.8%

97.0%

Snorer C

17

15

2

3

88.2%

83.3%

Snorer D

2646

2522

124

46

95.3%

98.2%

Snorer E

881

859

22

2

97.5%

99.8%

OSAS

Patients

OSAS A

1849

1545

304

284

83.6%

84.5%

OSAS B

475

466

9

70

98.1%

86.9%

OSAS C

731

725

6

47

99.2%

93.9%

OSAS D

2331

2326

5

5

99.8%

99.8%

OSAS E

1965

1904

61

184

96.9%

91.2%

10 patients

12338

11596

742

738

94.0%

94.0%

Table 3-2. System performance validation results of locating OSAS segments

Group

Patient

OSAS segments

TP

FN

FP

Sensitivity

PPV

OSAS

Patients

OSAS A

318

235

83

14

73.9%

94.4%

OSAS B

150

110

40

43

73.3%

71.9%

OSAS C

288

264

24

31

91.7%

89.5%

OSAS D

56

53

3

80

94.6%

39.8%

OSAS E

93

58

35

56

62.4%

50.9%

5 OSAS patients

923

720

185

224

79.6%

76.3%

Table 3-3. System performance validation results of classifying normal segments

Group

Patient

Normal segments

TP

FN

Sensitivity

Normal snorers

Snorer A

360

283

77

78.6%

Snorer B

360

319

41

88.6%

Snorer C

360

360

0

100%

Snorer D

360

316

44

87.8%

Snorer E

360

311

49

86.4%

5 normal snorers

1800

1589

211

88.3%

SOD can be a stand-alone device for snoring and OSAS symptoms detecting. It can also be combined with PAD-Mat for further OSAS diagnosis. The of OSAS symptoms detected by SOD, a pause of snoring sounds, can also be caused by body position change. Therefore when combined with PAD-Mat, a one-minute segment is recognized as a normal segment if body movements are also detected. Figure 3-13 shows the OSAS segments versus time analyzed by PSG and SOD of OSAS patient D and Figure 3-14 shows OSAS segments versus time analyzed by SOD combined with PAD-Mat. In this case, sensitivity dropped from 94.6% to 91.1%, but PPV was improved from 39.8% to 47.2%. Overall, the average sensitivity of locating OSAS segments dropped slightest from 79.6% to 77.0%, but the PPV improved from 76.3% to 80.1%. The lowest PPV (OSAS patient D) improved from 39.8% to 47.2% (Table 3-4). For normal group, the average sensitivity of recognizing normal segments was also improved from 88.3% to 89.8% (Table 3-5).

Figure 3-13. OSAS segments versus time (OSAS patient D)

Figure 3-14. Body movements and OSAS segments versus time (OSAS patient D)

Table 3-4. System performance validation results of locating OSAS segments

Group

Patient

OSAS segments

TP

FN

FP

Sensitivity

PPV

OSAS

Patients

OSAS A

318

230

88

5

72.3%

97.9%

OSAS B

150

107

43

36

71.3%

74.8%

OSAS C

288

251

37

28

87.2%

91.6%

OSAS D

56

51

5

57

91.1%

47.2%

OSAS E

111

58

35

47

62.4%

55.2%

5 OSAS patients

923

97

208

173

77.0%

80.1%

 

Table 3-5. System performance validation results of classifying normal segments

Group

Patient

Normal segments

TP

FN

Sensitivity

Normal snorers

Snorer A

360

293

67

81.4%

Snorer B

360

328

32

91.0%

Snorer C

360

360

0

100%

Snorer D

360

322

38

89.4%

Snorer E

360

314

46

87.2%

5 normal snorers

1800

1617

183

89.8%

3.5     Summary

OSAS is characterized by apnea during sleep usually following snoring. Recent research has shown that there is a high prevalence of OSAS in snorers and a high incidence of cardiovascular disease in OSAS patients [Lugaresi, et al., 1980, Hla, et al., 1994, Foley, 1999, Peker, et al., 1999, Peled, et al., 1999]. A portable device for tele-monitoring of snoring and OSAS symptoms has been developed in this research to evaluate snoring with quantitative measurement and detect OSAS symptoms with the W-ISR in real time. In our validation tests, the sensitivity and positive predictive value of detecting snoring are no different between normal snorers and OSAS patients. Furthermore, the W-ISR is a useful index to represent the character of intermittent snoring and achieved good sensitivities of locating OSAS segments and normal segments.

The SOD is a nonobtrusive user-centered device which allows users to easily set up a personal monitoring system at home and gain a better understanding of their sleep problems. As mentioned in Section 3-1, the SOD is one of the many applications of the Portable Telehomecare Monitoring System (PTMS). What sets the PTMS apart from most other systems is the focus on a highly decentralized monitoring model and the portable nature of the system. We believe that this is the approach that is needed to make such systems economically viable and acceptable to the end-users.

Reference

Foley, D.J., “Association of symptoms of sleep apnea with cardiovascular disease, cognitive impairment and mortality among older Japanese-American men”, Journal of the American Geriatrics Society, 1999, v. 47, n. 5, p.524-528.

Hla, K.M., Young, T.B., Bidwell, T., Palta, M., Skatrud, J.B., and Dempsey, J., “Sleep apnea and hypertension. A population-based study”, Annual of Internation Medicine, 1994, v. 120, n. 5, p.382-386.

Hoffstein, V., Mateika, S., Nash, S., “Comparing perceptios and measurements of snoring,” Sleep, v.19, n.10, 1996, p.783-789.

Lugaresi, E., Cirignotta, F., Coccagna, G., Piana, C., “Some epidemiological data on snoring and cardiocirculatory disturbances,” Sleep, v. 3, n. 3, 1980, p.221-224.

Peker, Y., Kraiczi, H., Hedner, J., Loth, S., Johansson, A., and Bende, M., “An independent association between obstructive sleep apnea and coronary artery disease”, European Respiratory Journal, 1999; v. 14, p.179-184.

Peled, N., Abinader, E.G., Pillar, G., Sharif, D., and Lavie, P., “Nocturnal ischemic events in patients with obstructive sleep apnea syndrome and ischemic heart disease”, Journal of the American College of Cardiology, 1999; v. 34, p.1744-1749

Penzel, T., Amend, G., Meinzer, K., Peter, J. H., von Wichert, P., “MESAM – A heart-rate and snoring recorder for detection of obstructive sleep-apnea,” Sleep, v.13, n.2, 1990, p.175-182.