DETECTION OF DAILY LIFESTYLE CHANGE FROM PULSE

RATE MEASURED DURING SLEEP

Wenxi Chen

, Hiroo Watanabe

, Xin Zhu

, Kei-ichiro Kitamura

and Tetsu Nemoto

Biomedical Information Technology Laboratory, the University of Aizu, Aizu-wakamatsu, Japan

Department of Laboratory Science, Faculty of Health Sciences, Institute of Medical, Pharmaceutical and Health Sciences

Kanazawa University, Kanazawa, Japan

Keywords: Daily lifestyle, Pulse rate, Sleep, Dynamic time warping (DTW).

Abstract: This study aims at detecting changes in daily lifestyle by using pulse rate measured during sleep. A

convenient system for pulse rate measurement during sleep and an algorithm for detection of lifestyle

changes were developed in this study. The data collection system consists of a home unit and a database

server. The home unit includes a Bluetooth-enabled SpO

sensor and a relay station. The sensor measures

pulse rate (PR) and SpO

beat-by-beat. The relay station receives the measured PR and SpO

data via

Bluetooth connection with the sensor, and then transmits these data to the database server through Internet

automatically. The database server manages the data and performs data analysis. Daily PR data were

preprocessed to suppress spike-like noise and movement artefact. Changes in daily lifestyle were detected

by a dynamic time warping (DTW) algorithm. Vital data were collected from a healthy college student

during daily sleep time over one year, and were used to examine the prototype system. The results showed

that unusual or irregular events, such as too much alcohol drink, physical illness and mental stress, could be

identified successfully. The system seems promising in application of health care and management under

daily life environment.

1 INTRODUCTION

Many chronic diseases, such as diabetes mellitus,

hypertension, angiosclerosis, hypercholesterolemia

and obesity, usually were conceived over a long

period accompanying with poor lifestyle in daily

living. Study showed that intensive lifestyle changes

may lead to regression of coronary atherosclerosis

after one year. More regression of coronary

atherosclerosis occurred after five years than after

one year (Ornish et al., 1998).

Positive changes in lifestyle are believed helpful

in dealing with chronic conditions. Changes in diet

and exercise are effective in curbing the

development of diabetes in older people (NIH news,

2006). An active lifestyle with appropriate and

sufficient physical activity was recommended

beneficial in intervening body weight for obesity

(Keim et al., 2004).

Monitoring of overt human behavior during

normal daily life has become feasible. One of the

most commonly used methods to assess daily

activities is based on an ambulatory accelerometry

(Welk et al., 2000); (Bussmann et al., 2001).

Further, a home-based system deploys a number

of low-cost sensors within the home to continually

monitor movement of older people, doors opening or

closing, environmental temperature and appliance

usage. By gathering these lifestyle data over a period

of time, a template is created as an average of the

client’s daily lifestyle profile. This template is then

used as a reference by which deviations of lifestyle

can be detected (Barnes et al., 1998).

AMON, a wearable medical monitoring and alert

system, continuously collects multiple vital signs,

detects emergent situation for high-risk cardiac and

respiratory patients by integrating the whole system

in an unobtrusive, wrist-worn watch without

interfering the patients daily activities and without

restricting their mobility (Anliker et al., 2004).

A sensor vest was developed to monitor multiple

vital signs, such as ECG, pulse, body temperature

and acceleration, to transmit data via mobile phone,

and to perform data analysis for early warning of

lifestyle-related diseases (Chen et al., 2005).

358

Chen W., Watanabe H., Zhu X., Kitamura K. and Nemoto T..

DETECTION OF DAILY LIFESTYLE CHANGE FROM PULSE RATE MEASURED DURING SLEEP.

DOI: 10.5220/0003738703580361

In Proceedings of the International Conference on Health Informatics (HEALTHINF-2012), pages 358-361

ISBN: 978-989-8425-88-1

 2012 SCITEPRESS (Science and Technology Publications, Lda.)

The purpose of this study is to develop an

automatic system for collecting pulse rate (PR)

conveniently during sleep, and an algorithm for

detecting changes in daily lifestyle. We propose a

method to define an initial template of average sleep

pattern from daily PR profiles and to evolve on daily

base. The template is used as a reference to calculate

its similarity with the daily PR profile by the

dynamic time warping (DTW) algorithm. We define

a sleep index (SI) based on the DTW result to

descrcibe daily lifestyle changes. The proposed

method is examined by one-year’s data collected

from a healthy college student.

2 METHOD AND MATERIAL

PR data were collected during sleep using the

scheme as showed in Figure 1. The system includes

a wrist-type Bluetooth-enabled SpO

sensor (Model

4100, Nonin Corp., USA), a bedside box and a

database server. Although the SpO

sensor collected

both PR and SpO

data simultaneously, only the PR

data were used in this study. The bedside box

receives PR and SpO

data continuously from the

SpO

sensor via the Bluetooth connection and

transmits the data to the database server by an HTTP

connection through a domestic LAN at home during

sleep. The database server served for data

management, analysis and visualization.

Figure 1: Schematic of PR and SpO

data collection

during sleep.

2.1 Data Collection

The bedside box is always on a standby status

waiting for the SpO

sensor to initiate. The

Bluetooth connection between the bedside box and

the SpO

sensor device will be established

automatically after the subject lies down to sleep and

inserts the SpO

sensor into a finger to trigger the

SpO

sensor switch. The bedside box receives the

measured PR and SpO

data from the SpO

sensor

via Bluetooth connection, and then transmits these

data to the database server through Internet once a

minute. When the subject gets up and removes the

sensor from the finger, the Bluetooth connection is

closed, the bedside box goes into standby mode

again, and the data collection procedure is

terminated.

The subject involved in data collection was given

the explanation of the task and study purpose, and

was asked to sign a consent agreement prior to the

data collection. The subject was a male college

student in twenties of age, and was allowed to live in

usual lifestyle completely over one-year period. The

subject was asked to insert the finger into the SpO

sensor every night as possible as he could, but

occasional skip in measurement in some nights due

to personal reason or temporary leave are permitted.

Information of the actual daily activities is collected

by subject’s diary.

2.2 Data Preprocessing

Data preprocessing for noise suppression was

implemented using two digital filters in two steps. A

median filter was used in the first step to remove

spike-like noise, and a Savitzky–Golay filter was

used in the second step to smooth the PR profile.

The main source of noise in the raw

measurement during sleep is spike-like noise, which

is perhaps due to movement artefacts, or

misinterpretation of the transmitted data package.

Such noise is suppressed in the first step.

The Savitzky–Golay filter was used to smooth

the signal that was outputted from the median filter

in the first step.

One night sample of PR profiles, raw and

preprocessed, are shown in Figure 2.

Figure 2: Pulse rate profile during sleep. Raw PR data

(thin black trace) and filtered PR data (bold red trace) in

single night’s sleep.

DETECTION OF DAILY LIFESTYLE CHANGE FROM PULSE RATE MEASURED DURING SLEEP

359

2.3 Detection of Lifestyle Change

Lifestyle change is detected by the dynamic time

warping (DTW) algorithm, which is used to measure

the similarity between two data sequences that may

generally vary in temporal span and rhythmic tempo.

A similarity measure for the reference pattern R

and the test pattern T is determined by the overall

minimum distance D(T, R). A smaller distance value

indicates a higher similarity.

A template of PR profile was initialized by

averaging five regular nights’ PR data measured

during sleep. Because daily data differs in length,

the data lengths in all five nights were normalized to

six hours. The initial template adapts evolution on

daily base gradually to reflect intrinsic biorhythmic

change.

The preprocessed daily PR profile was used in

the detection of daily lifestyle change. The template

of PR profile was used as the reference pattern, and

the daily PR profile was used as the test pattern.

The sleep index (SI) S is defined by normalizing

the similarity measure, or the overall minimum

distance between two patterns, D(T, R) as below:

(

)

(

)

(

)

min max min

,/SDTRD D D=- -

(1)

where the value S is between 0 and 1, D

min

and D

max

indicate the minimum and maximum similarity

values, respectively. The smaller the value S is, the

more regular the sleep is.

3 RESULTS

The results by analyzing daily PR data over one year

from a healthy male college student in his twenties

are presented in Figure 3. The value S was

calculated using the DTW method from daily PR

profile and the template of PR profile. The lower the

value S is, the higher the similarity is with usual

daily lifestyle. The higher values of days, whose

values are greater than 2SD, indicate the days when

the subject had a heavy intake of alcohol, gotten

illness or mental stress or other unusual events. It

can be observed that 2SD is a proper threshold for

detecting irregular lifestyle changes in the proposed

method.

4 DISCUSSIONS

Many lifestyle-related chronic diseases threaten

human beings. Better lifestyle is believed to be

significant in dealing with such diseases. A healthy

lifestyle requires a balanced nutrition diet, regular

physical activities and proper coping with mental

stress. Lifestyle improvement needs to persist.

Therefore, it is important to monitor routine lifestyle

and its change by a sustainable means.

A daily lifestyle can be considered consisting of

a series of physical and mental activities. A variety

of wearable devices, such as a wrist-watch or an

earphone-like device, an arm belt or a vest, for

daytime use usually impede routine activities more

or less. Convenient and comfortable monitoring

methods are rarely available. It is indispensible to

monitor daily lifestyle in a convenient way yet

sensitive enough to detect any lifestyle change

whenever occurs.

Physical and mental conditions are affected by

various endogenous and exogenous factors. The

former includes emotional, psychological, and

Figure 3: Variation in SI over one year period. Balloon boxes indicate some special events recorded by the subject, and

correspond to the higher SI values calculated by the proposed method.

HEALTHINF 2012 - International Conference on Health Informatics

360

behavioural aspects, and the latter includes

meteorological, environmental, geographical, and

temporal factors. Human body has to levy upon

immune system and auto regulation mechanism to

adapt various regular and irregular stimulants

physically and mentally. Activities in daytime often

preserve memories for a certain period

physiologically and psychologically, and usually can

be reflected as a physiological response at night.

On the other hand, sleep insufficiency or disorder

may cause unpleasantness even illness. Good sleep

can help to not only secrete growth hormone and

recover human body’s physiological functions, but

also relieve mental aspect from stress and build up

immune system.

Instead of identifying every detail of daily

activities in daytime, we believe that it is possible to

monitor physiological condition during sleep at night

to reflect the lifestyle change in daytime indirectly.

Standard polysomnography method provides an

accurate approach to monitor multiple parameters

and perform comprehensive sleep analysis, but

requires professional intervention and is highly

expensive, therefore is unsuitable for daily

application at home.

We developed a convenient device for automatic

collecting PR data during sleep and an algorithm to

detect lifestyle change from these PR data. Various

specific events, such as alcohol drink, mental

depression and physical illness, and other commonly

non-routine epochs in daytime are confirmed often

bringing disturbance or disorder in sleep at night,

and are probably reflected on night-time PR profiles.

This study demonstrated availability to detect these

daily behavioural changes during waking hours by

the PR data collected during sleep.

This method is recognized feasible for a user

over one year test. However, more data from more

users in different age groups and longer period of

data collection are desirable in further validation of

the proposed method. More sensitive and robust

algorithms are also worth to be explored in depth.

5 CONCLUSIONS

In this study, we developed a convenient system to

measure PR and SpO

data during sleep, and a

DTW-based algorithm to detect lifestyle change

using daily PR profile. The proposed method was

examined by one-year data and confirmed sensitive

in detecting lifestyle change due to various

incentives. It suggests a promising method for daily

health management and chronic disease prevention.

ACKNOWLEDGEMENTS

The authors would like to thank the volunteer for his

endurance in daily data collection over a long period.

REFERENCES

Ornish, D., Scherwitz, L., Billings, J., Gould, K., Merritt,

T., Sparler, S., Armstrong, W., Ports, T., Kirkeeide, R.,

Hogeboom, C., Brand, R., 1998. Intensive Lifestyle

Changes for Reversal of Coronary Heart Disease, The

Journal of the American Medical Association;

280(23):2001-2007

NIH News, Lifestyle Changes Can Help Older Hispanics

Manage Diabetes

http://www.nih.gov/news/pr/apr2006/nia-18.htm.

Accessed Aug. 1, 2011.

Bussmann, J., Martens, W., Tulen, J., Schasfoort, F., Berg-

Emons, H., Stam, H., 2001. Measuring daily behavior

using ambulatory accelerometry: The Activity Monitor,

Behavior Research Methods, Instruments, &

Computers; 33 (3): 349-356

Anliker, U., Ward, J., Lukowicz, P., Troster, G., Dolveck,

F., Baer, M., Keita, F., Schenker, E., Catarsi, F.,

Coluccini, L., Belardinelli, A., Shklarski, D., Alon, M.,

Hirt, E., Schmid, R., Vuskovic, M., 2004. AMON: a

wearable multiparameter medical monitoring and alert

system. IEEE Transactions on Information Technology

in Biomedicine; 8(4): 415–427

Chen, W., Wei, D., Ding, S., Cohen, M., Wang, H.,

Tokinoya, S. and Takeda, N. 2005. A Scalable Mobile

Phone-Based System for Multiple Vital Signs

Monitoring and Healthcare. International Journal of

Pervasive Computing and Communications; 1(2):157-

163

Barnes, N., Edwards, N., Rose, D., Garner, P., 1998.

Lifestyle monitoring-technology for supported

independence. Computing & Control Engineering

Journal; 9(4):169-174

Keim, N., Blanton, C., Kretsch, M., 2004. America’s

obesity epidemic: Measuring physical activity to

promote an active lifestyle. Journal of the American

Dietetic Association; 104(9):1398-1409

Welk, G., Blair, S., Wood, K., Jones, S., Thompson, R.,

2000. A comparative evaluation of three

accelerometry-based physical activity monitors.

Medicine & Science in Sports & Exercise; 32(9),

Suppl.:S489-S497.

DETECTION OF DAILY LIFESTYLE CHANGE FROM PULSE RATE MEASURED DURING SLEEP

361