6th International Special Topic Conference on ITAB, 2007, Tokyo

Bluetooth Based Wireless Sleep Remote Monitoring System

K.M. Chang', M. S.

Shihl, T. H. Lan', L. W. Hang2

'Department of Computer and Communication Engineering, Asia University, Taichung, Taiwan

2Chief of Respiratory Care unit, China Medical University Hospital, Taichung, Taiwan
Abstract- Sleep disorder, such as Obstructive sleeps
apnoea (OSA), and is now regarded as an important risk factor

monitor OSA patients for their homecare need. This system includes MSP430 chip as A/D converter, that is low-power requirement. ECG amplifier with Bluetooth transmission function is also included. This system can remote monitor the patient's ECG signal, that is useful for hospital medications to monitor patients sleep in time. An OSA state determination algorithm, based on the HRV LF power, can

for cardiovascular diseases. Traditional sleep monitoring

system contains several biomedical signals recording set, such as EEG, EOG, EMG, ECG, respiration and SpO2. The

increasing demands of home healthcare for sleep drive studies

to monitor sleep by simple method. This paper proposed an

ECG-based wireless sleep monitoring system, which contains a

Bluetooth-based ECG transmission system, and algorithms to

give a warning sign for the high OSA risk patients to visit
the doctors.
II. METHOD

monitor OSA by ECG-derived features, with clustering

sensitivity 77% and specificity 90%. This system is believed to

be a good candidate to meet the future demand on home-based

A. System structure

sleep monitoring, and it can reduce the abundant burdens of hospital sleep centers.

The system described in this paper has three major

components:

Keywords

Bluetooth, ECG, OSA, Sleep

I. INTRODUCTION

(1) Hardware:
We develop an ECG amplifier with gain 100 and low

pass filter within 40 Hz. The A/D converter is made by the

OSA is one of the most physiologic disruptive and

dangerous of sleep-related diagnoses, and it has been increasingly implicated in the initiation and progression of cardiovascular diseases [1]. The standard approach to diagnosis of OSA is with an in-laboratory, technician-attended full Polysomnography (PSG), but it's labor-intensive and time-consuming. Therefore an alternative approach, such as portable monitoring for home diagnosis is a hot research topic [2]-[3]. In 2000, Computers in Cardiology (CINC) and PhysioNet encourage investigating sleep monitoring by ECG, and many ECG-derived parameters may be useful candidate [4]. In this paper, we setup a low-power ECG monitoring system, with wireless transmission by Bluetooth, and implemented an OSA diagnosis algorithm based on ECG to

MSP430F449 IC, produced by TI instruments. This IC has a low power requirement and compatible with 8 channel

analog input. Four AAA cell can provide the system power efficiently. This property can make the system smaller and being portable. The MSP430 sampling frequency is 200 Hz. (2) Bluetooth ECG transmission: There are several wireless transmission methods, such as Bluetooth, Zigbee, RF, and WiFi. The Bluetooth transmission rate is 2.1 Mbps, operated on 2.4GHz, transmission range from 10 cm to 100 meter. We choose Bluetooth due to its short distance, high transmission capacity, and higher security concern. Low power requirement is also a major factor. All such Bluetooth properties are good for home-based ECG transmission. We adopt the Bluetooth module from J3 Company [5],

978-1-4244-1868-8/08/$25.00 (C2008 IEEE.

309

6th International Special Topic Conference on ITAB, 2007, Tokyo

and combined with MSP430 chip and ECG amplifier within

an IC board. Bluetooth transmit ECG signal to the host

Chinese Medical Hospital, Taichung, Taiwan. The threshold

RDI= 15 is chosen because patients with RDI> 15 are judged as light sleep disorder [6].

computer.

(3) Host PC monitoring interface and remote monitoring: The ECG monitoring interface is programmed by
D. OSA identification algorithm

Labview. We also program a remote monitoring interface beside the host computer that is also coded by Labview. Figure 1 is the illustration of the system structure described above.

There are five steps of the proposed OSA

identification algorithm:
(1) Choose five hour sleep ECG from the whole night recording.

(2) R peak detection by Tompkin' s algorithm.

(3) Extract time domain and frequency domain heart rate interval features. That includes mean heart rate, RMSSD,
SDNN, LF and HF power [7]. We also divide the HRV
spectrum with 0.02 Hz interval as features.

(4) Statistical testing by t-test analysis. (5) OSA clustering performance by receiver operating characteristic (ROC) curve. All the above algorithms are implemented by Matlab, and coded into the Labview module.
Fig. 1. Structure of Bluetooth_based sleep ECG monitoring system
IV. RESULTS

B. ECG transmission performance testing

A. Hardware system

The ECG transmission performance of the system had been tested by three aspects. At first, we compare the
ECG amplifier performance with commercial ECG amplifier.

The ECG amplifier output is shown on the scope, as in

Second, the digital output ECG signal was compared with analog ECG to test the MSP430 A/D converter performance. The test ECG signals are from simulator and real ECG signals. Third, the Bluetooth transmission performance; include the ECG signal loss after transmission and transmission range are also investigated.
C. OSA subjects information

There are sixty subjects, include thirty patients with

respiratory disturbance index (RDI)>15 (26 male and 4 female, age 42.8+10.7, and RDI is 77.4+15.5, range from 103.8 to 41.5), and another thirty patients with RDI<15 (19 male and 11 female, age 35.2+12.5, and RDI is 6.9+4.4,
range from 13.8 to 0), data are recorded from sleep center of

figure 2a. The digital ECG pattern after MSP430 are recorded by Labview through RS233 connect; and the further Bluetooth transmitted ECG are also compared, and both shown in figure 2b. The likelihood between figure2b is tested by template matching method. Maximum value is 0.96, that means the ECG output between MSP430 (before Bluetooth transmission) and Bluetooth transmission ECG is very similar. Bluetooth transmission doesn't distort ECG pattern. Figure 3 is the user interface coded by Labview. The remote monitoring site has similar interface. In the interface, the ECG can be saved and reload for further analysis. Figure 2b is also the demonstration of the reloaded ECG signal,
which is saved on the host computer after Bluetooth transmission.

978-1-4244-1868-8/08/$25.00 (C2008 IEEE.

310

6th International Special Topic Conference on ITAB, 2007, Tokyo

Fig. 2a. Analog output of the ECG amplifier on the scope.

Fig. 4. R wave peak detection result (marked in circle)

Table 1.
The ECG-derived parameters for RDI> 15 and RDI< 15.
r

C1
'I

DLl~l

flfD

2t

S [I0SI4,D IIfl

4fD

Ut
--I

parameters

RDl<15

RDI>15
18.214.1 *E8

P value

HRV Power

7.43.8 *E8

0.0002

.Q.. --.
a

(0.04-0.06 Hz)
----

LB_.rL.

10 fl5

O(

20X

X5

flD

IDO

power(0.04-0.06
Hz )/ total power

0.236 0.084

0.377 0.129

7.69E-06

Fig. 2b

LF/(LF+HF)
LF power

0.5710.148
19.510.7 *E8

0.6370.135
36.432.4 *E8

0.006
0.010

LF/HF

1.7221.368

2.6681.724

0.022

8000000000 7000000000

im 6000000000
N 5000000000
o 4000000000

4 3000000000
2000000000

Fig. 3. User interface coded by Labview.

1000000000
10 15 20 25 30

B. OSA identification algorithm

Fig. 5. Distribution of HRV power between 0.02 to 0.04 Hz.

wave

Figure 4 is the illustration of R

peak detection
can

Figure 6 is the corresponding ROC

specificity 90
1.1 1
0.9

curve

of figure 5. The
and

result. The original five-hour long sleep ECG

be

optimal clustering performance is sensitivity 77

00.

processed to identify the R peak position (which is marked

in circle) within 42 seconds by PC (1.8 G Hz CPU and 480
MB RAM) operated in Window XP.

The extracted RR interval is fed into further analysis. Table

1

list the ECG derived parameters for both

power

group,

cm 0.6 * 0.5
= 0.4

4'

0.8 0.7

RDI>15 and RDI <15, and their t-test results. The most

significant feature is the HRV spectrum

The OSA patients
are

between

0.3 0.2
0.1

0 0

0.02-0.04 Hz. Figure 5 is the corresponding distribution.

0.2

0.4

0.6

0.8

marked

square

and the normal

Another
user

1-sensitivity Fig. 6. ROC performances

subjects are marked diamond.

interface implemented by Labview and

978-1-4244-1868-8/08/$25.00 (C2008 IEEE.

311

6th International Special Topic Conference on ITAB, 2007, Tokyo

used for OSA estimation, is also constructed and shown in

The OSA identification algorithm proposed in this paper can also used with conventional CPAP system to monitor the

figure 7. Patients can upload their sleep ECG into the system,
and less than one min later, the interface will give different

sleep performance of OSA patients.

This system takes the advantages of the wireless and

color light for different sleep performance. If the patient had

a bad sleep, the red light will turn on to notice the patient

internet, doctors and patient' s relatives can monitor the

takes a further examination in the hospital. The analysis result is also connected to the hospital database, and the doctor can trace the patient's OSA treatment performance immediately.

patient sleep and to take a proper medical treatment in time. It would be very useful in the future,
ACKNOWLEDGMENT

This work has been supported by the National Science

Council of Taiwan (grant number 95-2218-E-468-002-).
REFERENCES

[1] J. F. Pagel, "Obstructive Sleep Apnea (OSA) in Primary Care:

Evidence-based Practice," The Journal of the American Board of Family

Medicine. 20 (4): 392-398, 2007.

[2] W. Ward Flemons,etc, " Home Diagnosis of Sleep Apnea:A Systematic

Fig. 7

Review of the Literature An Evidence Review Cosponsored by the American Academy of Sleep Medicine, the American Collegeof Chest

V. DISCUSSION and CONCLUSION

Physicians, and the American.Thoracic Society," CHEST 2003;

124:1543-1579.

In this paper, we propose a system contains one-lead

ECG, which can be transmitted wireless by Bluetooth system.

[3] S. J. Redmond and C. Heneghan, "Cardiorespiratory-Based Sleep

The host computer contain signal demonstration and user interface. HRV sub-band power based OSA identification

Staging in Subjects with Obstructive Sleep Apnea," IEEE Trans. Biomed.

Eng., vol. 53, no.3. pp.485-496, MARCH 2006.

algorithm is also implemented in the system. This system is based on the whole night ECG to estimate the OSA. Traditional R wave detection algorithm developed on Tompkins is very useful in this study. The major goal of this study is to develop homecare oriented sleep monitoring system, and ECG is chosen due to its vital sign characteristics. Low cost, low power demand, easy to setup and friendly user interface are all major factor to be considered. Such system is especially useful for the primary screen on OSA candidate and for further traces after OSA
treatment.

[4] T. Penzel, etc, "Systematic comparison of different algorithms for

apnoea detection based on electrocardiogram recordings," Med. Biol. Eng.

Comput., 2002, 40, 402-407.

[5] BT102x-BT20-Module- User's Guide
-

[6] M. H. Kryger, Principles And Practice Of Sleep Medicine, W. C.

Dement, Ed. Elsevier Science Health Science div, 2000.

[7] R. M. Rangayyan, Biomedical Signal Analysis, JOHN WILEY, 2002.

Bluetooth combined with MSP430 is especially useful

to meet the biomedical instrumentation requirement. Low

power, portable, high transmission capacity, high security,

and practical transmission range is the major advantages.

978-1-4244-1868-8/08/$25.00 (C2008 IEEE.

312