Development of a Portable Device for Tele-monitoring of Physical Activities During Sleep

Chih-Ming Cheng, Ph.D.,1 Yeh-Liang Hsu, Ph.D.,1 and Chang-Ming Young, M.D.2

1 Department of Mechanical Engineering, Yuan Ze University, Chung-Li, Taiwan, ROC.

2Ming Young Biomedical Corp., Junan, Miaoli, Taiwan, ROC.

AbstractLow motor activity levels and prolonged episodes of uninterrupted immobility are characteristics of sleep. In clinical practice, the use of polysomnographic (PSG) recording is a standard procedure to assess sleep. However, PSG is not suitable for long-term monitoring in the home environment. This paper describes the development of a portable telemonitoring device that detects movements of a subject by conductive mats, and evaluates sleep stages via physical activity data. The device itself also serves as a Web server. Doctors and caregiv-ers can access real-time and historical data via an IE browser or a remote application program for telemonitoring of physical activities and sleep/awake states during sleep, while the patients stay in their own homes. In our validation test with four normal subjects and four arousal subjects, this system showed a good performance in locating sleep epochs of a subject. The sensitivity of locating sleep epochs was 89.5% and the average positive prediction value was 94.8%, with a specificity of 84.3%. This device is not intended to be a diagnosis device, instead, it is to be used as a home telehealth tool for monitor-ing physical activity and sleep/awake states. This portable telemoni-toring device provides a convenient approach to better understand and recognize a subject’s sleep pattern through long-term sleep monitoring in the home environment.

Key words: home telehealth, physical activity, sleep pattern

Introductionuring sleep, low motor activity levels and prolonged episodes of uninterrupted immobility are associated with increasing sleep depth. There are also motor disturbances that are trig-gered by sleep such as restless legs syndrome (RLS) and peri-

odic limb movements during sleep (PLMS). Such symptoms disrupt sleep and cause daytime tiredness and sleepiness.

Several noninvasive and unrestrained sensing techniques have been developed for the monitoring of body movement during sleep. The use of load cells or force sensors is the most common approach to detect body movements in bed. Load cells represent a simple and durable technology, which was used by several researchers.1–4 The use of force sensor is also a popular technique for monitoring body movements in bed. Nishida et al.5 presented the idea of a robotic bed, which is equipped with 221 pressure sensors for monitoring of respiration and body position. Van der Loos et al.6 also proposed a similar system called SleepSmart™, composed of a mattress pad with 54 force sensitive resistors and 54 resistive temperature devices, to estimate body center of mass and index of restlessness. This large-size equipment is difficult to set up and can only be used in specific laboratories.

Many pad-based solutions have been proposed. Several authors have used the static charge sensitive bed (SCSB) for monitoring motor activity. The SCSB is composed of two metal plates with a wooden plate in the middle that must be placed under a special foam plastic mattress, which operates like a capacitor.7–9 Watanabe et al.10 designed a pneumatics-based system for sleep monitoring. A thin, air-sealed cushion is placed under the bed mattress of the subject and the small movements attributable to human automatic vital functions are measured as changes in pressure using a pressure sensor.

Other sensing techniques, such as optical fibers and conductive fibers have also been used for monitoring body movement in bed. Tamura et al.11 proposed a body movement monitoring system using

1044 TELEMEDICINE and e-HEALTH DECEMBER 2008 DOI: 10.1089/tmj.2008.0026

O R I G I N A L R E S E A R C H

Development of a Portable Device for Telemonitoring of Physical Activities During Sleep

D

© MARY ANN LIEBERT, INC. • VOL. 14 NO. 10 • DECEMBER 2008 TELEMEDICINE and e-HEALTH 1045

TELEMONITORING OF PHYSICAL ACTIVITIES DURING SLEEP

optical fibers. Kimura et al.12 designed an unobtrusive vital signs detection system, which uses conductive fiber sensors to detect body position, respiration, and heart rate. These fiber sensors can be incor-porated in a conventional bed sheet for home use.

Polysomnography (PSG) is considered to be the gold standard method for assessing sleep. Different sleep stages are evaluated by medical specialists using PSG data such as electroencephalogram (EEG), electrooculogram (EOG), and electromyogram (EMG) based on the Rechtschaffen and Kales (R-K) method. This technique usually requires individuals to sleep in research laboratories that are known to change habitual sleep patterns13 and induce a first-night effect.14

Since the development of the actioculographic monitor in 1979, a novel method to estimate sleep stage through body movements have been suggested. Measurement of motility has become a popular method in the study of human sleep. Kayed et al.15 used three criteria, (1) eye movements, (2) body movements, and (3) submental electromyogram, to identify wake, rapid eye movement (REM) sleep, and non-rapid eye movement (NREM) sleep. Based on this development, a wearable wrist actigraphy, has been developed for the identification of wake, REM, and NREM stages.16–18 Ajilore et al.19 proposed a home-based sleep monitor-ing system, called Nightcap, which uses eyelid and body movement sensors to discriminate awake, NREM, and REM sleep automatically. The literature shows that actigraphy is a valuable device to detect sleep–wake period. The agreement between actigraphy and PSG with epoch-to-epoch comparison, ranged from 80% to 90%.16,20 Although estimat-ing sleep stages through body movement is not as accurate as PSG, studies are in general agreement that measures with body movement were fairly sensitive in detecting sleep.

Similar to wrist actigraphy, Choi et al.21 introduces bed actigraphy for user-friendly sleep–wake monitoring. An automatic scoring algorithm scores each epoch of the recordings for either “wake” or “sleep” by the signals of four load cells, which are installed at each corner of the bed. Bed actigraphy has some advantages over ordinary actigraphic devices.

First, bed actigraphy does not constrain the subjects during record-ing. The subject can even be unaware of the recording process. Second, bed actigraphy can detect the overall movements of the sub-ject by measuring the body activity at four points, while an actigraph detects only the movements of the body part on which it is mounted. The bed actigraphy device developed by Choi et al.21 showed good agreement (95.2%) but medium sensitivity (64.4%) and positive pre-dictive value (66.8%).

For the needs of remote monitoring of sleep, several research activities are underway. Kristo et al.22 presented a telemedicine pro-tocol for the online transfer of PSGs from a remote site to a central-ized sleep laboratory, which provided a cost-saving approach for the diagnosis of obstructive sleep apnea syndrome (OSAS). Seo et al.23 developed a nonintrusive health-monitoring house system to monitor patients’ electrocardiogram results, weight, movement pattern, and snoring in order to get the information of patients’ health status and sleep problems. Choi et al.24 presented a ubiquitous health monitor-ing system in a bedroom, which monitors ECG, body movements and snoring with nonconscious sensors.

A centralized framework is used in most home telehealth systems, in which a centralized database is used for data storage and analysis. Figure 1 illustrates the structure of the decentralized home telehealth system developed by the authors.25 Instead of using a centralized database that gathers data from many households, a single household

Fig. 1. The structure of the decentralized home telehealth system.

Java applet server

Browser

VB program

E-mail box/mobile phone

Household Care-giver/Client user

Application server

Applet

Eventmessage

Eventalert

Regularreport

Cable/wirelesstransmitter

Cable/wirelessreceiver

PIC server

MMCDistributed data server

Request/Data

E-mail server/SMS

Sensor 3

Sensor 2

Sensor 1

1046 TELEMEDICINE and e-HEALTH DECEMBER 2008

CHENG ET AL.

is the fundamental unit for sensing, data transmission, storage and analysis. The core of the system is the Distributed Data Server (DDS) inside a household, which is a thin server designed specifically for the decentralized home telehealth system. It consists of a PIC server mounted on a peripheral application board. The PIC server integrates a PIC microcontroller (PIC18F6722, Microchip Technology, Inc., Chandler, AZ), EEPROM (24LC1025, Microchip) and a networking IC (RTL8019AS, Realtek, Taiwan, ROC). The DDS provides networking capability and can be used as a Web server.

As shown in Figure 1, sensing data from sensors embedded in the home environment are transmitted to the DDS, and are then processed and stored in the Multi-Media-Card (MMC) of the DDS. Authorized remote users can request data from the DDS using a Web browser (running Java applets) or a Visual Basic (VB) program. Event-driven messages (mobile phone text messages or e-mails) can be sent to specified caregivers when an urgent situation is detected.

There are several advantages of the decentralized structure over the traditional centralized database structure:

1. The scale of the home telehealth system is much smaller, which makes it economically viable and acceptable to the end-users.

2. Instead of sending the health monitoring data to a centralized database in a home healthcare provider, health monitoring data are stored within the household. Only authorized caregiv-ers can access the data. Privacy is better protected.

3. The route from the sensor to server is much shorter. Data trans-mission is easier and more reliable. When the Internet commu-nication fails, the local system can still function normally and keep collecting data. Thus, data integrity is better preserved.

4. The home telehealth provider only needs to maintain an application server to provide email, short message, and Java services. The DDS can also be used as a gateway if a central-ized database is needed.

This paper describes the development of a Physical Activity Detecting Mat (PAD-Mat) based on the decentralized home telehealth system structure described above. Figure 2 shows the structure of the PAD-Mat developed in this research. Similar to the decentralized structure in Figure 1, the core component of the PAD-Mat is a DDS that detects activity signals from upper limb, body and leg, captured by three conductive mats. Physicians and caregivers can access the DDS for real-time and historical data via an IE browser or a remote application program for telemonitoring of physical activities and sleep/awake states during sleep, while the patients stay in their own homes. If the subject in bed has no activities for a predefined period of time, an event message (mobile phone short messages or emails) can be sent to specified caregivers.

The PAD-Mat provides a nonconstrained and nonconscious approach for telemonitoring of physical activities during sleep. It can be used for in-bed detection and monitoring motor disturbances, such

as RLS and PLMS. In addi-tion, using the sleep activity index (SAI) proposed in this research, sleep and wakeful stages can be classified on-line. This portable device is not intended to be a diagnosis device, instead, it is to be used as a home telehealth tool for monitoring physical activity and sleep/awake states.

DESIGN OF THE PAD-MATAs shown in Figure 2, three

conductive mats placed under the chest, hip, and legs are used in the PAD-Mat to detect the physical activities with the resistance changes of the mats. The signals are analyzed with the physical activity detecting Fig. 2. Structure of the Physical Activity Detecting Mat (PAD-Mat).

Java applet server

Browser

VB program

E-mail box/mobile phone

Household Care-giver/Client user

Application server

Applet

Eventmessage

Eventalert

Regularreport

PIC server

MMC

Distributed data server

Request/Data

E-mail server/SMS

Cable

Upper LimbConductiveMat

LegConductiveMat

BodyConductiveMat

© MARY ANN LIEBERT, INC. • VOL. 14 NO. 10 • DECEMBER 2008 TELEMEDICINE and e-HEALTH 1047

TELEMONITORING OF PHYSICAL ACTIVITIES DURING SLEEP

algorithms. The following sections describe the design of the conduc-tive mats, the in-bed detecting algorithm, the physical activity evaluat-ing algorithm and the sleep/awake identifying algorithm in details.

DESIGN OF THE CONDUCTIVE MATSThe conductive mats are made of conductive fabric (Ming Young

Biomedical Corp., Taiwan, ROC). There are two types of conduc-

tive mat as shown in Figures 3 and 4. The PAD-Mat comprises two Type-A mats for detecting upper limb and leg activities (the upper limb mat and the leg mat) and one Type-B mat for detect-ing body activities (the body mat). Physical activities are cap-tured as the electrical resistance changes with motion on each mat. The construction of the conductive mat is inspired by the intestinal villi, which increase the surface area of absorption. Sensitivity of this bended structure is significantly higher than that of a straight structure. A layer of nonconducting foam is inserted in the Type-B mats so that the electrical resistance of the Type-B conductive mats can sufficiently restore after the subject leaves the bed. A simple dividing circuit is used to converts resis-tance changes of the body mat into voltage changes. Figure 5shows the placement of the PAD-Mat on the bed.

The analog signals are digitized at a sampling frequency of 2 KHz with a 10-bit A/D converter in the PIC server. Figure 6A shows a 16-minute record of the body mat input signals. The subject changed body position at around the second, eighth, and thirteenth minute. A series of smoothing and simplifying procedures further processes these signals. A moving average filter, with a window size of 20, smoothes the profile of the input voltage values. After that, the signals are further simplified into 10 data points per second to save computational resource in the following detecting algorithms (Fig. 6B). This procedure is applied to signals from all three conduc-tive mats.

IN-BED DETECTING ALGORITHMThe body mat is also used to detect whether the subject is in

bed or not. The input voltage decreased when the subject lies on

Fig. 3. (A) Appearance of conductive mats (Type-A). (B) Dimensions of conductive mats (Type-A).

Fig. 4. (A) Appearance of conductive mats (Type-B). (B) Dimensions of conductive mats (Type-B). Fig. 5. Placement of the Physical Activity Detecting Mat (PAD-Mat).

(A)

(A)

(B) 1 cm 1 cm

1 cm

(B) 1 cm 1 cm 1 cm 1 cm 1 cm 1 cm

1 cm1 cm

3 cm

FOAM FOAM

Upper Limb Mat(Type A Conductive Mat)

Body Mat(Type B Conductive Mat)

Leg Mat(Type A Conductive Mat)

1048 TELEMEDICINE and e-HEALTH DECEMBER 2008

CHENG ET AL.

the mat, and stays at 3.3 V after the s ubject leaves the mat. With a threshold limit 3.2 V, data are coded “1” if below the threshold limit and a “0” if the slope value is above the threshold limit. Figure 7 shows a one-night record of the input voltage of the body mat. The subject went to bed at around 2:35 AM, and left bed at around 8:15 AM. Body movements might cause transient increases of voltage that results in a rapid switch from “0” to “1.” To avoid that problem, the PAD-Mat uses 1 minute as the time unit to define the “In-bed Code.” If there are more “1” than “0” in that minute, the In-bed Code of the minute is coded “1” and vice versa. As shown in Figure 7, the PAD-Mat recognizes the input voltage values as a series of In-bed Codes and the total in-bed time is 340 minutes.

PHYSICAL ACTIVITY DETECTING ALGORITHMAfter the smoothing and simplify-ing procedures, the physical activ-ity detecting algorithm analyzes the data and output activity indexes by the following steps:1. Calculate the absolute slope val-

ues of activity signals in Figure 6B, as shown in Figure 8A.

2. Symbolize a slope value with an activity code “1” if it is above the threshold limit and a “0” if the slope value is under the threshold limit. Figure 8B shows a 10-second clip of Figure 8A, a body movement was recognized as a series of Activity Codes “1,” using a threshold limit 0.025. The threshold limit for upper limb and leg activities are both 0.02.

Fig. 6. (A) Input voltages of body activities. (B) Smoothed and simplified values of the input voltage.

Fig. 7. In-bed detection by the body mat.

2:00 3:00 4:00 5:00 6:00 7:00 8:00 9:00

Time (hr:min)

1.5

2.0

2.5

3.0

3.5

Inpu

t vo

ltage

(Vo

lt)

2:00 3:00 4:00 5:00 6:00 7:00 8:00 9:00

Time (hr:min)

0

2.0

2.5

3.0

1

In-b

ed c

ode

2.4

2.6

2.8

3.0

0:00 1:00 2:00 3:00 4:00 5:00 6:00 7:00 8:00 9:00 10:00 11:00 12:00 13:00 14:00 15:00 16:00

Time (min:sec)

2.4

2.6

2.8

3.0

0:00 1:00 2:00 3:00 4:00 5:00 6:00 7:00 8:00 9:00 10:00 11:00 12:00 13:00 14:00 15:00 16:00

Time (min:sec)

(B)

(A)

© MARY ANN LIEBERT, INC. • VOL. 14 NO. 10 • DECEMBER 2008 TELEMEDICINE and e-HEALTH 1049

TELEMONITORING OF PHYSICAL ACTIVITIES DURING SLEEP

3. Tally the counts of “1” series each minute as the Body Activity Index (BAI), which represents the length of activ-ities occurred in the past min-ute. Figure 8C shows the body activity index versus time of Figure 6A. Three body move-ments, which lasted 1.5, 1.2, and 1.5 seconds, have been recognized.

Following the same procedures, the PAD-Mat also tallies the Upper Limb Activity Index (ULAI) and the Leg Activity Index (LAI) for each minute. Figure 9 shows a 7.5-hour record of the ULAI, BAI, and LAI of a subject evaluated by PAD-Mat. ULAI, LAI, and BAI, which represent the duration of activities occurring in a minute, can be used for the monitoring of motor disturbances that are triggered by sleep such as RLS and PLMS. For example, over-lapping the figures for ULAI, BAI, and LAI in Figure 9, it can be found that a portion of the movements only occurs at the upper limb area or the leg area. These indexes provide quantitative information on the fre-quency of the movements, but not the magnitudes of the movements.

THE SLEEP/AWAKE DISTINGUISHING ALGORITHM

In sleep, motor activity is reduced in comparison to the waking state. The frequency of all movements decreases with depth of sleep, with progressive decrease in the number of movements from sleep stage I to stage IV.26

Since the late 1970s, a growing number of studies have demon-strated the validity of actigraphy in

Fig. 8. (A) Absolute slope values of body movement of a subject (16 minutes). (B) Activity code of a body movement (threshold limit value = 0.02). (C) The body activity index versus time.

0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16

Time (min)

Slop

e va

lue

8:15 8:16 8:17 8:18 8:19 8:20 8:21 8:22 8:23 8:24 8:25

Time (min:sec)

Slop

e va

lue

Activity code 0000000000000000000000000000000000000000000111111111111000000000000000000000000000000000000000000000

Body

act

ivity

inde

x

0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16

Time (min)

0.00

0.02

0.04

0.06

0.08

0.10

0.000

0.025

0.050

0.075

0.100

5.04.54.03.53.02.52.01.51.00.50.0

(A)

(B)

(C)

1050 TELEMEDICINE and e-HEALTH DECEMBER 2008

CHENG ET AL.

distinguishing between sleep and wakefulness. Jean-Louis et al.27 uti-lized a simple technique for scoring sleep/awake epochs. The epochs containing higher activity than a given threshold were scored as wake and lower activity as sleep, and arousals lasting 3 minutes or less were rescored as sleep. Cole et al.28 proposed two rules for identi-fying sleep and awake epochs: (1) the sleep of 1, 3, or 4 minutes was rescored as wake if it preceded at least 4, 10, or 15 minutes of wake, respectively and (2) the sleep of 6 or 10 minutes surrounded by at least 10 or 20 minutes of wake was rescored as wake, respectively.

Sadeh et al.29 used logistic regression analysis for the variables of body activity while the sleep/awake classification of PSG acted as the dependent variable in the analysis. Five activity variables were computed for each epoch in the activity signals: original value, mean, standard deviation, number of epochs above a specified activity level, and the natural logarithm. Most actigraphy-related devices scored signals using similar algorithms to classify sleep/awake epochs, based on variables of body activity.

Rachwalski et al.30 used a pressure pad placed below a person’s hips to measure activity in bed. Activity in bed is measured by changes in the pressure measurements. The pressure measurements aggregated

into 30-second epochs by aver-aging the data every 30 seconds. Awakenings (or periods of restless-ness) are defined by 3 consecutive minutes of body movements. Choi et al.21 designed a bed actigraphy sys-tem for distinguishing between sleep and awake. Signals were coded as “1” as the intensity of signals higher than the threshold. If the duration of “1” is longer than 3 seconds in an epoch (30 second length), the epoch is scored as “Wake.”

In summary, both the quantity and tendency of physical activity are considered as the character of physical activity and used for the evaluating of level of wakefulness. To identify the subject’s sleep con-dition in real time, ULAI, BAI, and LAI described in the previous section are used to recognize whether the subject is asleep in the PAD-Mat. To consider the tendency of activ-ity, weighted activity indexes are

calculated by considering the activity indexes of the past minutes with different weighting, as shown in equation 1. For example, weighted-BAI (W-BAI) of the fifth minute is calculated as equation 2. Weighted-ULAI and weighted-LAI (W-LAI) are also calculated with the same equation.

k k

Weighted – BAIi = Σ (k–n+1).BAIi–n+1 /Σn (1) n=1 n=1

(5.BAI5 + 4.BAI4 + 3.BAI3 + 2.BAI2 + 1.BAI)Weighted – BAI5 = ———————————————————————————————————————— ——————— (2)

15

To explain the relationship between physical activities and sleep depth, sleep reports of PSG and physical activity data of the PAD-Mat were collected from five subjects recruited specifically for this purpose. A sleep value (SV) was defined based on the PSG report and scored by minute. The SV of the minute is coded “0” if PSG recognized two epochs (1 minute) as awake, and “1” as two sleep epochs were detected. If there were only one sleep epoch in the

Fig. 9. Activity indexes of a subject from 0:30 to 8:00 AM.

0:30 1:00 1:30 2:00 2:30 3:00 3:30 4:00 4:30 5:00 5:30 6:00 6:30 7:00 7:30 8:00

Time (hour:min)

ULAI

0

2

4

6

8

10

BAI

0

2

4

6

8

10

LAI

0

2

4

6

8

10

© MARY ANN LIEBERT, INC. • VOL. 14 NO. 10 • DECEMBER 2008 TELEMEDICINE and e-HEALTH 1051

TELEMONITORING OF PHYSICAL ACTIVITIES DURING SLEEP

past minute, the SV was coded “0.5.” Regression analyses were used to explain the relationship between physical activities and SV. Regression analyses were used on three 80-minute epochs ran-domly selected from each subject, with weighted activity indexes and different k values in equation 1. For example, k = 5 means the

regression analysis model includes weighted-ULAI, weighted-BAI, and weighted-LAI, which were calculated using the raw data of the past 5 minutes.

Table 1 shows that result of regression analyses performed with different k values. ULAI was found to be nonsignificant with different k values, which means ULAI is not a valuable for sleep/awake clas-sifying. Furthermore, BAI and LAI are both highly significant with the k value of 5. Equation 3 is the regression equation used in this research, in which BAI and LAI were considered for evaluating sleep with the k value of 5.

SV = 0.659–0.028 W-BAI–0.026 W-LAI (3)

Figure 10 shows an 8-hour recording of weighted-BAI, weighted-LAI, and SV calculated with equations 1, 2, and 3.

Figure 11 shows an 8-hour recording (0:00 to 8:00 AM) of In-Bed Codes and Sleep Codes of the same data in Figure 10, as well as the final output “Sleep State.” In this research, the Sleep Code of the past minute is assigned as “1” if SV of this minute calculated by equation 3 is higher than 0.5, and “0” if SV is under the threshold limit 0.5.

Table 1. Regression Analyses with Different κ Values

κ VALUE

SIGNIFICANT

W-ULAI W-BAI W-LAI

1 N (p = 0.919) N (p = 0.195) N (p = 0.214)

2 N (p = 0.125) N (p = 0.559) Y (p = 0.032)

3 N (p = 0.082) N (p = 0.085) Y (p = 0.079)

4 N (p = 0.087) Y (p = 0.010) N (p = 0.258)

5 N (p = 0.390) Y (p = 0.008) Y (p = 0.002)

6 N (p = 0.165) Y (p = 0.049) Y (p = 0.019)

W-ULAI, weighted Upper Limb Activity Index; W-BAI, weighted Body Activity Index; W-LAI, weighted Leg Activity Index.

Fig. 10. Weighted Body Activity Index (W-BAI), weighted Leg Activity Index (W-LAI), and Sleep Value of a subject from 0:00 to 8:00 AM.

0 1 2 3 4 5 6 7 8

Time (hour)

W-BAI

Threshold

W-LAI

Sleep value

02468

10121416

02468

10121416

0.00.10.20.30.40.50.60.7

1052 TELEMEDICINE and e-HEALTH DECEMBER 2008

CHENG ET AL.

As shown in Figure 11, Sleep States were coded “2” (sleep) if both Sleep Code and In-bed Code are “1,” and “0” (Empty Bed) if both Sleep Code and In-Bed Code are “0.” Sleep State “1” (Awake) is recognized as the subject is in bed but stayed awake (In-bed Code 1 and Sleep Code 0). According to the output of the PAD-Mat in Figure 11, this subject went to bed at about 0:34 AM and fell asleep at 0:40 AM. After a 6-hour sleep, the subject woke up at about 6:50 AM and left bed at 6:50 AM.

USER INTERFACE AND NETWORK FRAMEWORK

Figure 12 shows the network framework for the PAD-Mat. Doctors and caregivers can access real-time and historical data via a remote VB program. Figure 13 shows the VB interface window designed for the PAD-Mat. After typing an IP address and clicking the real-time button, real-time information of physical activities per minute and sleep state of the subject are displayed on the left side of the interface window. Three sleep states, sleep, awake, and empty bed, will be identified.

Doctors and caregivers can monitor three patients on the same window. Selecting one of the IP addresses on the left side, the user can also download historical data from the PAD-Mat at this IP address (select a date and time interval on the right side of the interface win-dow and click the “download” but-ton). Monitoring reports for time to lie on bed and fall asleep, as well as total in-bed time and sleep duration within the selected time interval can be tallied automatically. Charts Fig. 12. A framework for the Physical Activity Detecting Mat (PAD-Mat).

Fig. 11. In-bed Code, Sleep Code, and Sleep States evaluated by the Physical Activity Detecting Mat (PAD-Mat).

0:00 1:00 2:00 3:00 4:00 5:00 6:00 7:00 8:00

Time (hour:min)

Sleep code

0

1

In-bed code

0

1

Sleep state

(empty bed) 0

(awake) 1

(sleep) 2

Doctor

Care givers

SMTP INTERNET GSM

House 1

House 2

House 3

PAD-Mat

PAD-Mat

PAD-Mat

PAD-Mat

PAD-Mat

PAD-Mat

House N

House N-1

House N+1

© MARY ANN LIEBERT, INC. • VOL. 14 NO. 10 • DECEMBER 2008 TELEMEDICINE and e-HEALTH 1053

TELEMONITORING OF PHYSICAL ACTIVITIES DURING SLEEP

of activity indexes of upper limb, body, and leg per minute versus time, and sleep states versus time can also be plotted. In addition to online monitoring, doc-tors and caregivers can also download the data from the DDS into a local PC for further off-line analysis. The DDS can also be used as a gateway for transmitting monitoring data to a centralized server for large-scaled data management.

In addition to motor disturbances mon-itoring and sleep states assessment, the PAD-Mat can be used for detecting body position changes. Bed sores are common problems in nursing homes. Usually care-givers have to change the subject’s body position every 2 hours. Figure 14 shows a 3-day history of a subject in a nursing home. Caregivers and family members can access the PAD-Mat from the Internet for tracking body activities of the patients.

VALIDATION TEST OF THE PAD-MATA validation test was designed to eval-

uate the performance of PAD-Mat. Eight subjects were recruited for the validation test. All eight subjects were young males, aged 20 to 25, with no sleep-related prob-lems. Eight subjects were divided into two groups. The four subjects in group 1 were allowed to sleep overnight, while the four subjects in group 2 were asked to leave bed and go to the restroom after they were woken by an alarm clock in the middle of sleep.

The PAD-Mat was used to detect sleep epochs of eight subjects in a full-night’s sleep. Sleep states judged by the PAD-Mat with Sleep Codes and In-bed Codes, were compared epoch-to-epoch with the moni-toring reports by a PSG. Figure 15 and Figure 16 show the sleep states scored via PAD-Mat and PSG of one subject in the first group and one subject in the second group. As shown in the figures, the sub- Fig. 14. Three-day history of a subject in a nursing home.

Fig. 13. Virtual Basic (VB) interface window for the Physical Activity Detecting Mat (PAD-Mat).

00 01 02 03 04 050 06 07 08 09 10 11 12 13 14 15 16 17 18 19 20 21 22 23 00

Time (hour)

BAI

2007/07/07

2007/07/09

2007/07/10

0

10

20

30

10

20

30

10

20

30

BAI

0

BAI

0

1054 TELEMEDICINE and e-HEALTH DECEMBER 2008

CHENG ET AL.

ject of group 1 did not leave bed until 7 AM, and the subject of group 2 went to the bathroom at around 5 AM for about 4 minutes. In both figures, the PAD-Mat seems to be more sensitive and recorded more awake episodes than PSG.

Table 2 shows the comparison results of locating sleep epochs using the PAD-Mat and PSG. A 1-minute epoch was coded true-positive (TP) if the PAD-Mat and PSG both classified as sleep epoch, and true-negative (TN) if the PAD-Mat and PSG both classified as non-sleep epoch (awake and empty bed). Similarly, a segment was coded false-positive (FP) or false-negative (FN) if only the PAD-

Mat or only the PSG recognized the state as sleep. Furthermore, in Table 2, sensitivity is defined as TP/(TP+FN), the proportion of sleep epochs correctly identified by the PAD-Mat and the total sleep epochs identified by PSG. The positive predictive value (PPV) is defined as TP/(TP+FP), which refers to the proportion of sleep epochs correctly identified by the PAD-Mat and the total sleep epochs identified by the PAD-Mat. The specificity is a statistical measure of how well a binary classification test correctly identifies the negative cases, and is defined as TN/(FP+TN). In our valida-tion test, the sensitivity of locating sleep epochs was 89.5% and

the average PPV was 94.8%, with a specificity of 84.3%.

CONCLUSIONSSince the invention of actigra-

phy, a growing number of studies have demonstrated how physical activities during sleep can be used in distinguishing between sleep and wakefulness. This is also a popular method for designing sleep monitor-ing devices used in the home envi-ronment.

In this research, the PAD-Mat, a portable device for telemonitoring of physical activities was developed to evaluate body movements with quantitative measurement and rec-ognize sleep, awake, and empty bed states. In our validation tests, the PAD-Mat shows good sensitivity and specificity in identifying sleep states, and can be a nonconstrained approach to better understand the sleep pattern.

One commonly used measure of sleep quality is the Pittsburgh Sleep Quality Index (PSQI), which evaluates sleep quality with seven “component” scores: subjective sleep quality, sleep latency, sleep dura-tion, habitual sleep efficiency, sleep disturbances, use of sleeping medi-cation, and daytime dysfunction. The sum of scores for these seven

Fig. 16. Sleep states scored via Physical Activity Detecting Mat (PAD-Mat) and polysomnographic (PSG; Group 2).

Fig. 15. Sleep states scored via Physical Activity Detecting Mat (PAD-Mat) and polysomnographic (PSG; Group 1).

0 1 2 3 4 5 6 7 8

Time (hour)

Sleep state(PSG)

Total sleep time351 minutes

0

1

2

0

1

2Sleep state(PAD-Mat)

Total sleep time318 minutes

0 1 2 3 4 5 6 7 8

Time (hour)

Sleep state(PSG)

Total sleep time494 minutes

0

1

2

0

1

2 Total

Total

Sleep state(PAD-Mat)

Total sleep time446 minutes

© MARY ANN LIEBERT, INC. • VOL. 14 NO. 10 • DECEMBER 2008 TELEMEDICINE and e-HEALTH 1055

TELEMONITORING OF PHYSICAL ACTIVITIES DURING SLEEP

components yields one global score of 1–21. Table 3 shows the rules for scoring sleep latency, sleep duration, sleep efficiency, and sleep disturbance of PSQI. The three objective components in PSQI, namely sleep latency, sleep duration, and habitual sleep efficiency, can be estimated through the asleep/awake detection of PAD-Mat. The two common sleep disturbances, RLS and PLMS, can also be monitored by the PAD-Mat.

The PAD-Mat system is designed for long-term home-based moni-toring on the decentralized home telehealth system structure. What sets the structure 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.

This device is not intended to be a diagnostic device, instead, it is to be used as a home telehealth tool for monitoring physical activity and sleep/awake states. This portable telemonitoring device provides a convenient approach to better understand and recognize a subject’s sleep pattern through long-term sleep monitoring in the home environment.

Future research will focus on the commercialization of the system. More user trials are to be conducted to evaluate the user acceptance, clinical effect, and economic analysis of the system. Other possible applications, elderly care in nursing homes for example, are to be explored in the future.

Disclosure StatementNo competing financial interests exist.

REFERENCES

1. Wheatley DW, Berme N, Ferguson-Pell MW. A low-profile load transducer for monitoring movement. Exp Mech 1980;20:19–20.

2. Barbenel JC, Ferguson-Pell MW, Beale AQ. Monitoring the mobility of patients in bed. Med Biol Eng Comp 1985;23:466–468.

3. Choi JM, Choi BH, Seo JW, Sohn RH, Ryu MS, Yi W, Park KS. A System for Ubiquitous Health Monitoring in the Bedroom via a Bluetooth Network and Wireless LAN. Engineering in Medicine and Biology Society, 2004. In EMBC 2004. Conference Proceedings. 26th Annual International Conference of the IEEE, 2004;2:3362–3365.

4. Adami AM, Hayes TL, Pavel M, Singer CM. Detection and classification of move-ments in bed using load cells. In IEEE 27th Annual International Conference of the Engineering in Medicine and Biology Society, 2005, pp. 589–592.

5. Nishida Y, Takeda M, Mori T, Mizoguchi H, Sato T. Monitoring patient respiration and posture using human symbiosis system. In Proceedings of the 1997 IEEE International Conference on Intelligent Robots and Systems, 1997;2:632–639.

6. Van der Loos HF, Ullrich N, Kobayashi H. Development of sensate and robotic bed technologies for vital signs monitoring and sleep quality improvement. Auton Robots 2004;15:67–79.

7. Alihanka J, Vaahtoranta K. A static charge sensitive bed—A new method for recording body movements during sleep. Electroencephalogr Clin Neurophysiol 1979;46:731–734.

Table 3. Components of PSQI

COMPONENT

SCORE

0 1 2 3

Sleep latency ■ 15 min 16–30 min 31–60 min > 60 min

Total sleep time > 480 min 360–480 min 300–360 min < 300 min

Sleep efficiency > 85% 75%–84% 65%–74% < 65%

Sleep disturbances 0 1–9 10–18 19–27

PSQI, Pittsburgh Sleep Quality Index.

Table 2. System Performance Validation Results of Locating Sleep Segments

GROUP ID SLEEP TIME TP FN FP SENSITIVITY PPV NON-SLEEP TIME TN SPECIFICITY

1

A 372 261 28 8 90.3% 97.0% 83 75 90.4%

B 438 275 42 20 86.8% 93.2% 121 101 83.5%

C 402 296 25 2 92.2% 99.3% 81 79 97.5%

D 554 320 40 39 88.9% 89.1% 194 155 79.9%

2

A 622 426 68 20 86.2% 95.5% 128 108 84.4%

B 399 283 8 32 97.3% 89.8% 108 76 70.4%

C 479 319 68 13 82.4% 96.1% 92 79 85.9%

D 267 259 8 0 97.0% 100% 70 66 94.3%

8 subjects 3533 2439 287 134 89.5% 94.8% 877 739 84.3%

1056 TELEMEDICINE and e-HEALTH DECEMBER 2008

CHENG ET AL.

8. Rauhala E, Erkinjuntti M, Polo O. Detection of periodic leg movements with a static-charge-sensitive bed. J Sleep Res 1996;5:246–250.

9. Kaartinen J, Kuhlman I, Peura P. Long-term monitoring of movements in bed and their relation to subjective sleep quality. Sleep Hypnosis 2003;5:145–153.

10. Watanabe K, Watanabe T, Watanabe H, Ando H, Ishikawa T, Kobayashi K. Noninvasive measurement of heartbeat, respiration, snoring and body move-ments of a subject in bed via a pneumatic method. IEEE Trans Biomed Eng 2005;52:2100–2107.

11. Tamura T, Nishigaichi A, Nomura T. Monitoring of body movement during sleep in bed. In Proceedings of the Annual International Conference of the IEEE Engineering in Medicine and Biology Society, 1992;14:1483–1484.

12. Kimura H, Kobayashi H, Kawabata K, Van der Loos HF. Development of an unobtrusive vital signs detection system using conductive fiber sensors. In Proceedings 2004 IEEE/RSJ International Conference on Intelligent Robots and Systems, 2004;1:307–312.

13. Reynolds III CF, Grochocinski VJ, Monk TH, Buysse DJ, Giles DE, Coble PA, Matzzie JV, Doman J, Monahan J, Kupfer DJ. Concordance between habitual sleep times and laboratory recording schedules. Sleep 1992;15:571–575.

14. Riedel BW, Winfield C, Lichstein KL. Anxiety and first night effect in older insomniacs. Sleep 1998;21:131.

15. Kayed K, Hesla PE, Rosjo O. The actioculographic monitor of sleep. Sleep 1979;2:253–260.

16. Hauri PJ, Wisbey J. Wrist actigraphy in insomnia. Sleep 1992;15:293–301.

17. Emilia S, Monica Z, Christophe P, Jean K. Actigraphy and leg movements during sleep: A validation study. J Clin Neurophysiol 1999;16:154–160.

18. Ancoli-Israel S, Cole R, Alessi C, Chambers M, Moorcroft W, Pollak CP. The role of actigraphy in the study of sleep and circadian rhythms. Sleep 2003;26:342–392.

19. Ajilore O, Stickgold R, Rittenhouse CD, Hobson JA. Nightcap: Laboratory and home-based evaluation of a portable sleep monitor. Psychophysiology 1995;32:92–98.

20. Sadeh A, Hauri PJ, Kripke DF, Lavie P. The role of actigraphy in the evaluation of sleep disorders. Sleep 1995;18:288–302.

21. Choi BH, Seo JW, Choi JM, Shin HB, Lee JY, Jeong DU, Park KS. Non-constraining sleep/wake monitoring system using bed actigraphy. Med Biol Eng Comput 2007;45:107–114.

22. Kristo DA, Eliasson AH, Poropatich RK, Netzer CM, Bradley JP, Loube DI, Netzer NC. Telemedicine in the sleep laboratory: Feasibility and economic advantages of polysomnograms transferred online. Telemed J E Health 2001;7:219–224.

23. Seo J, Choi J, Choi B, Jeong DU, Park K. The development of a nonintrusive home-based physiologic signal measurement system. Telemed J E Health 2005;11:487–495.

24. Choi JM, Choi BH, Seo JW, Sohn RH, Ryu MS, Yi W, Park KS. A system for ubiquitous health monitoring in the bedroom via a Bluetooth network and wireless LAN. In Proceedings of the IEEE 26th Annual International Conference on Engineering in Medicine and Biology Society, 2004;2:3362–3365.

25. Hsu YL, Yang CC, Tsai TC, Cheng CM, Wu CH. Development of a decentral-ized home telehealth monitoring system. Telemed E Health 2007;13:69–78.

26. Chokroverty S, Hening WA, Walters AS. Sleep and movement disorders. Philadelphia: Elsevier Science, 2003.

27. Jean-Louis G, von Gizycki H, Zizi F, Fookson J, Spielman A, Nunes J. Determination of sleep and wakefulness with the Actigraph Data Analysis Software (ADAS). Sleep 1996;19:739–743.

28. Cole RJ, Kripke DF, Gruen W, Mullaney DJ, Gillin JC. Automatic sleep/wake identification from wrist activity. Sleep 1992;15:461–469.

29. Sadeh A, Hauri PJ, Kripke DF, Lavie P. The role of actigraphy in the evalua-tion of sleep disorders. Sleep 1995;18:288–302.

30. Rachwalski T, Irvine S, Steeper JWT, Inkster DR, Wells C. Poor quality of sleep is a precursor of mobility-related adverse events. In Proceedings of Evidence-based strategies for Patient Falls and Wandering 2005, 2005.

Address reprint requests to:Yeh-Liang Hsu, Ph.D.

Department of Mechanical EngineeringYuan Ze University

135 Yuan Tung RoadChung Li 32003

TaiwanRepublic of China

E-mail: mehsu@saturn.yzu.edu.tw

Received: February 26, 2008Accepted: April 21, 2008