30th Annual International IEEE EMBS Conference Vancouver, British Columbia, Canada, August 20-24, 2008

Level of Activity, Night and Day Alternation, and well being measured in a Smart Hospital Suite N.Noury, Senior Member, IEEE, T.Hadidi, M.Laila, A.Fleury, C.Villemazet, V.Rialle, A.Franco

Abstract— The present paper reports a study on the daily activity of elderly people in a hospital suite, with presence infrared sensors. It is an attempt to produce parameters and indicators for the predictive analysis of the daily activity of fragile persons. A relationship is proposed between well being of the patient and the night and day activities alternation.

Sensors (PIR) were placed on the walls (height 1,80m) so as to detect the presence on the bed, the armchair, at the bath sink, in the cabinet and at the main entrance door (Fig 1).

I. INTRODUCTION

D

to the elderly boom, the geriatric institutions are facing a large scale problem. Beside, more and more people wish to live independently in their home, as long as it is possible. Therefore, a lot of researches has been done on the concept of “Health Smart Homes” [1-7] in order to provide the enabling technologies, sensors and communication networks to detect and report all the possible scenarios of dangerous situations that an elderly could face when living alone at home. The existing systems are mostly based on presence sensors and door contacts as these are affordable, reliable and easily accepted in the intimacy of the home; there are also some attempts to transpose directly the methods of scene analysis developed for the video monitoring of public places. Somehow, the major scientific challenge is not in the technologies and methods for sensing and transmitting the scenarios of emergency situations, but it is rather in the prediction of abnormal trends. The present paper reports a study of the daily activity of elderly, measured with presence detectors, with an attempt to produce the best parameters and methods for the predictive analysis of daily activity of fragile persons.

Sink

UE

II. MATERIAL AND METHODS A. The hospital suite A system based on the Health Smart Home concept, was built at the TIMC-IMAG Laboratory several years ago [8], namely the HIS. It is currently under test in a couple of real flats, but also in hospital suites [9]. The experimental set up described in this work, was installed in a hospital suite consisting of a main room, with bed and armchair, and a toilet. The Presence Infrared Manuscript received April 7, 2008. This work was supported in part by the French ministry of research under Grant n°03B651-9. Authors are with the TIMC-IMAG Laboratory, University of Grenoble, France (corresponding author N.Noury: 33476637486; e-mail: Norbert.Noury@ imag.fr).

978-1-4244-1815-2/08/$25.00 ©2008 IEEE.

Entrance

Cabinet

Armchair

Bed

Figure 1. The experimental set up is a Hospital suite. The 5 PIRs are monitoring continuously the activities of the patient and visits of the health professionals.

B. Data processing The raw data consists in a list of events of detection [DateTime-Localization] automatically written in a log file in XML format, which is transmitted once a day by Email to a central data base [10]. Upon reception, the file is loaded in the Matlab™ environment, where data is stored in a preliminary matrix built with detection events (1). Detection[Date][Hour][Sensor Number] with Date[Year][Month][Day] Hour[H][M][S] Sensor Identifier є [1,5]

(1)

Data is then formatted for further computations: time is converted into a discrete time dimension i , in seconds. The date is replaced by a day number j , the elapsed day from the beginning of the observation. For further digital signal processing, the data is completed with zeros where no detections were recorded. The signal S(j,i) (2) describes sampled detections at regular instants. n=S(j,i) (2) with j є [start day, end day] i є [1,86400] : seconds n є [0,5] : 1 to 5 = sensor number, 0 = no detection

3328

From the signal S(j,i) we compute a first parameter of “movements” from counting the events during a given time. But the time representation of S(j,i) shows discontinuities. Therefore we can hardly read or display it as is. Hence we apply a “fisrt order sample & Hold” which maintains the last detection valid until a new detection occurs at a different sensor. This non-linear filtering thus converts ‘detection of movements’ in front of the sensors into a ‘signal of presence’. Actually, this process only reduces the complexity of data with no impact on the mean time at each sensor as the subject is redetected when leaving to the next room. The final S*(j,i) signal (3) results from the application of this transformation on S(j,i). We call this process "rectangularization". n=S*(j,i) with j є [start day, end day] i є [1,86400] : seconds n є [0,5] : 1 to 5 = sensor number

(3)

The new graphical representation of the daily signal SJ*(i)= S*(J,i) , which we call “ambulatogram” (Figure 2), shows immediately the main periods of activity, and inactivity,

• • • • •

the mean number of events for each sensor, the mean time occupancy for each room, the number of diurnal events the total number of nocturnal events the activity profile (time-frequency representation)

The resulting data is a new file with, for each day of data, the number of events for each sensor (4), [Day][nocturnal events][entrance][bed][armchair][sink][cabinet] [mean number of displacements] [mean number of movements] (4)

As an example of real data we can find, [May 21] [1301] [11][33][0][1][0][21.92][184.38] Our goal is to produce relevant indicators to point out the non visible trends in the activity in the suite, so as to inform the Health team, in charge of the patient, with potential abnormal trends. Therefore, we decided to focus on the following indicators: • the mean activities in front of each sensor (indicator of movements), • the mean displacements from one sensor to another (indicator of displacements), • the mean activity during the day (diurnal activity) and • the mean activity at night time (nocturnal activity). Each indicator is normalized for the whole period of observation. For example, for indicator number i, namely indi, and the value vali on day number j, we obtain (5), Indi(j) = vali(j)/ Max{vali} With i = identifier of the indicator and j=day.

(5)

As an example we can obtain the processed data, [May 21] [0.70] [0.13][0.48][0][1][0][0.26][0.47] Figure 2 - The “ambulatogram” SJ*(i) represents the detections in each room with the time of day.

A further analysis allows the interpretation of the activities and occupations (sleeping time, nursing, meals). Several parameters are easily computed from this representation: • the total time spent in each place is obtained with cumulating the length of each time segments for each location • the number of events (“displacements”) during a given time comes from counting the vertical lines • the parameter “movements” has been previously defined from signal S(j,i) in equation (2). C. Parameters and Indicators From this raw data, several parameters were thus elaborated [11], such as:

After normalization all the data are in the same space of data [0…1]. III. RESULTS A.

Experimentation The data was collected during a 2 months period (May 21 to July 18th, 2007) in the hospital suite. The occupant was an elderly women aged 86, suffering from a congenital dysplasy of the 2 hips but she had good and painless spontaneous mobility in the bed and with the wheel chair. Although polypathologic, she had a stable chronic state but with multiple and recurring somatic complaints, without an organic cause.

3329

during night and day continuously).

B.

Correlation between movements and displacements For this patient, we first observe a strong correlation between the Diurnal displacements and movements. It means that this person is equally active in all the places (Fig 3).

0,60

0,40

0,20

1,00

Displacements Distance

0,90

Movements 0,80

0,00

12/5

22/5

1/6

11/6

21/6

1/7

11/7

21/7

-0,20

Normalised Values

0,70

0,60

-0,40

0,50 -0,60

0,40

0,30

-0,80

Date 0,20

Figure 5- Euclidian Distance between diurnal and night activity. Monitoring during a 2 months period (May-July 2007)

0,10

0,00

12/5

22/5

1/6

11/6

21/6

1/7

11/7

21/7

Date

Figure 3. Displacements and movements during a 2 months period (MayJuly 2007). The diurnal displacements and movements are strongly correlated.

A high rate of activity corresponds to periods of great complaining from the patient (i.e. on the figure: June 6th and July 12th). A low rate of activity may correspond to external events (on June 24th the patient was out on a permission for the week end).

Thus we tried to find out a representation of the phase shift between night and day activities, which we obtained by computing the trends (first derivate) computed on both signal on a day-to-day time step (Fig. 6). On this representation the “abnormal” synchronisation of nights and days is straightforward: when both the trends are either positive, or negative, it shows a synchronisation. 0,80

Nocturnal Trend

C. Correlation between diurnal and nocturnal activities The second observation is that the night activity is most of the time correlated to the daily activity (Fig.4 & 5). However, from time to time, when the night activity is higher, the next diurnal activity is lower. As probably the person feels tired after having a bad sleep at night.

0,60

Diurnal Trend 0,40

1st Derivate

0,20

0,00

12/5

22/5

1/6

11/6

21/6

1/7

11/7

21/7

-0,20

-0,40 1,20

Nocturnal events Movements (Diurnal)

-0,60

Normalised Number of events

1,00

-0,80

Date

Figure 6. The diurnal and nocturnal trends are correlated. Monitoring during a 2 months period (May-July 2007).

0,80

0,60

After questioning the physician on the written observations about this patient, we discovered that the periods of synchronization corresponded to periods when the patient felt poorly, complaining a lot to the medical team.

0,40

0,20

0,00

17/5

27/5

6/6

16/6

26/6

6/7

16/7

26/7

IV. CONCLUSION

Date

Figure 4. The diurnal and nocturnal activities are correlated. Monitoring during a 2 months period (May-July 2007).

After questioning the physician, he exposed us the following basic rules: • Following an agitated day, the subject normally sleeps well (low night activity) • Following an agitated night, the subject is sleepy all day (low daily activity) • Other behaviours are abnormal (low, or high, activity

The diurnal activity parameters can be evaluated with a simple set of PIR sensors distributed in the functional zones in the place of living, either a home or a hospital suite. There are multiple, and complex, relationships between the parameters of activity, the health status and the well being of the person. We have found visual correlations between night and day activities and trends. But those results, based on 1 subject only, only constitute a case study and we wait for more

3330

results with more subjects before to make general statements. This research aims at the automatic assessment of the health status of a person living independently in home. Therefore it could be possible to timely produce this information to the physician, or the relatives, in charge with the person. Eventually, it could make the elderly person feel more secure at home, and then this is an enabling technology to support the concept of “Aging in place”. ACKNOWLEDGMENT The AILISA project was supported by the French RNTS Health Network from 2003-07 within the framework of the “Institut de la Longévité” (n°03B651-9). The authors also wish to thank the scientific network “GDR-STIC-Santé” for his support. REFERENCES [1]

N. Noury, G. Virone, P. Barralon, J. Ye, V. Rialle, J.Demongeot, “New Trends in Health Smart Homes”, in Healthcom2003, SantaMonica-Califonia, 6-7 Jun 2003, pp 118-127 [2] A. Yamaguchi, M. Ogawa, T. Tamura, T. Togawa, “Monitoring behaviour in the home using positioning sensors”, in proc. 20th IEEE conference on engineering in Medicine and Biology, pp 1977-79, 1998. [3] M. Ogawa and T. Togawa, "Attempt at Monitoring Healt Status in the Home," in Proc. 1st Int IEEE-EMBS Special Topics Conf. on Microtechnol. in Med. and Biol., A. Dittmar and D. Beebe, Eds. Lyon, France: IEEE, 2000, pp. 552-556. [4] S.J. Richardson, D.F. Poulson, C. Nicolle, “Supporting independent living through adaptable smart home (ASH) technologies”, in Human welfare and technologies: papers from the human service information technology applications (HUSITA) conference on information technology and the quality of life and services, pp 87-95, 1993. [5] B.G. Celler, W. Earnshaw, E.D. Ilsar, L. Betbeder-Matibet, M.F. Harris, R. Clark, T. Hesketh, N.H. Lovell, “Remote monitoring of the health status of the elderly at home, a multidisciplinary project on aging at the University of New South Wales”, International journal of Biomedical Computing, 40:147-155, 1995. [6] M. Chan, C. Hariton, P. Ringeard, E. Campo, “Smart House automation system for the elderly and the disabled”, in IEEE International conference on Systems, Man and Cybernetics, pp15861589, 1995. [7] A.R. Kaushik, B.G. Celler “Characterization of PIR detector for monitoring occupancy patterns and functional health status of elderly people living alone at home”, Tech. and healthcare, vol 15 (4), 2007, pp.273-288. [8] N. Noury, T. Hervé, V. Rialle, G. Virone, E. Mercier, Monitoring behavior in home using a smart fall sensor and position sensors, in IEEE-EMBS « Microtechnologies in Medicine & Biology », LyonFrance, Oct 2000, pp 607-610 [9] N. Noury et al. « AILISA : plateformes d’évaluations pour des technologies de télésurveillance médicale et d’assistance en gérontologie », Journal Gérontologie et Société , n°113, Juin 2005, pp. 97-119. [10] N. Noury , C. Villemazet, A. Fleury, P. Barralon, P. Rumeau, N. Vuillerme, R. Baghai, “Ambient Multi-Perceptive System with Electronic Mails for a Residential Health Monitoring System”, in 28th IEEE-EMBC2006, New-York, 31 Aout - 3 sept. 2006, pp. 3612-15 [11] G. Le Bellego, N. Noury, G. Virone, M. Mousseau, J. Demongeot, “Measurement and Model of the Activity of a Patient in his Hospital Suite”, IEEE Trans. on Information Technology in Biomedicine, vol.10, Jan.2006, pp.92-99

3331