Testing of a long-term fall detection system incorporated into a custom vest for the elderly. 1.Department of Electronic and Computer Engineering, University.

Testing of a long-term fall detection system incorporated into a custom vest for the elderly.

1. Department of Electronic and Computer Engineering, University of Limerick, Limerick, Ireland

2. Department of Electronic Engineering, National University of Ireland, Galway, Ireland.

3. NCBES, National University of Ireland, Galway, Ireland.4. University of Pittsburgh, Pittsburgh, U.S.A.

Alan K Bourke1,3 ,Pepijn WJ van de Ven1,

Amy E. Chaya4, Gearóid Ó Laighin2,3, John Nelson1

Falls and the long-lieFalls and the long-lieFalls in the elderly are a major international health concernOne in every three community dwelling elderly experience

at least one fall every year. (Nevitt et al.,1989,JAMA)A serious consequence of a fall is the so-called ‘long-lie’

(Lord et al., 2001) which is defined as remaining on the ground, following a fall, for longer than one hour (Wild et al., 1981,BMJ)Approx 50% of those elderly who lie on the floor for an hour or more

die within 6 months. (Wild et al.,1981,BMJ)The solution to the problem of the long-lie:

Automatic detection of falls followed by transmission of an alert to the emergency services or care-giver

Current Autonomous long-term Fall-detection systems

Previously the end of a fall is characterized by an impact shock and/or the person being in a horizontal orientation

A number of long-term autonomous long-term fall-detection systems currently exist

Doughty et al., 2000 Jnl. of Telemedicine and Telecare(Tunstall group)

Noury et al., 2004, EMBCIncorporated into a vest

Karantonis et al., 2006, IEEE Trans. Info. Technol. Biomed.Worn at the waist Fall sensor by Noury et al.

Fall sensor

Tunstall group fall detector

Karantonis et al. fall and activity monitor

The Fall-sensorA fall-sensor was

developed which consists of:μController Bluetooth moduleTri-axial Accelerometer

(MMA7261)BatteryμSD card

The Vest Developed using feedback from elderly

subjects using questionnaire on donning, wearing the vest for an hour, then doffing the vest

The fall sensor can be located at the CHEST or LEFT UNDER-ARM, attached using Velcro

Zips closed at the front for ease of donning and doffing

100% polyester which is easy to clean, durable and stretchable

Additional elastic incorporated to support the sensor

Features of the Vest The vest is made fully

adjustable at the sides to fit subjects from 2XS to 3XL

Minimum amount of Fabric is located around the shoulders to allow for freedom of movement

Elastic was only present on one location where the sensor was to be located

Fall sensor

ObjectivesObjectives1. Test the light-weight sensor to implement the

fall detection algorithm2. Gain user feedback on the design of the

suitability of the vest to be worn by the elderly3. Evaluate the threshold based algorithm to

automatically distinguish between fall events and ADL.

StudiesStudies Two separate studies were completed:

1. A simulated fall-event and ADL study - used to establish the thresholds that would indicate that a fall had occurred. This was carried out using 10 young healthy subjects performing falls and normal activities

2. A long term Activities of Daily Living (ADL) study - to determine the accuracy of the algorithm in a long term trial. This was carried out using 10 elderly (>65 years) volunteers in a nursing home over the course of 4 weeks.

Study 1 - Falls and ADL with Study 1 - Falls and ADL with Young healthy subjectsYoung healthy subjects

This was carried out using 10 young healthy subjects performing falls and normal ADL while wearing the vest and attached fall sensor

A total of 8 different fall types were performed

Falls in all four direction, performed with both legs straight and knee flexion

A total of 5 different ADL were performed which included

sitting, lying, standing and walking activities

Fall Detection Algorithm Study 1 Fall Detection Algorithm Study 1 resultsresultsThe shaded algorithm was tested

against the recorded data

Following this subject posture is continually monitored forfall-alert: Lying for >75% of

10minutesfall-recovery: Lying for < 75% of

10minutes

Sensitivity Specificity Accuracy

Chest 91.3% 99.6% 94.4%

LU arm 97.1% 98.6% 97.7%

Study 2 - Long-term ADL trial with Elderly Study 2 - Long-term ADL trial with Elderly subjectssubjects

A total of 10 elderly subjects wearing the vest and fall-sensor over the course of 4 weeks.

The 10 elderly were divided into 2 teams Each team wearing the system for 2 week each for 8 hours a day from

Monday to Sunday. The trials took place in the nursing home “Benincasa” in the city of

Ancona, Italy and were coordinated by COOSS Marche Onlus

Team 1 Team 2

Week 1 CAALYX Fall system (Chest) rest

Week 2 rest CAALYX Fall system (Chest)

Week 3 CAALYX Fall sensor (Left under-arm) rest

Week 4 rest CAALYX Fall system (Left under-arm)

MethodMethodSubjects donned the system in the morning and then

proceeded to carry out their normal daily routine which included:Sitting transitionsLying transitionsWalkingTravelling by bus and carDiningTaking the stairsUsing the elevator

MethodMethodMessages from the fall-sensor

were relayed to the care-centre using Bluetooth to the Nokia N95 (which was attached to the subject in a pocket on the vest)

Messages were relayed further via the 3G network and the internet.

Messages includedFall-eventFall-alertFall-recovery

Fall-message transmission algorithmFall-message transmission algorithmFall-messages associated with falls are sent by the fall-sensor to the

mobile-phone and further propagated to the care-centreAcknowledgement of receipt of the fall-messages are sent from the

mobile phone to the fall-sensor and from the care-centre to the mobile-phone.

An acknowledgement of receipt of the fall-message from the mobile-phone to the care-centre is also sent back to the fall-sensor

No indication that this procedure is in progress is relayed to the elderly subject thus achieving automatic independent fall detection.

MethodMethodThe following messages thus appeared at the care-taker

terminal:Fall-eventFall-alertFall-recovery

The relevant steps were then taken by the care-staff to ensure the elderly subjects safety.

These messages were also logged by the fall-sensor along with the raw accelerometer data to the μSD card

Care-taker site

Results – Fall AlgorithmResults – Fall Algorithm In total 833 hours of monitoring was recorded over the course of 4

weeks onto the SD cardsFall-events

115 true fall-events were recorded and transmitted by the fall-sensor 144 true fall-events were registered at the care-taker site

Fall-alerts 42 fall-alerts were recorded and transmitted by the fall-sensor 9 fall-alerts were registered at the care taker site

Fall-recoveries 73 fall-recoveries were recorded and transmitted by the fall-sensor 52 fall-recoveries were registered at the care taker site

The discrepancy between the fall-messages received and those registered is due Bluetooth transmission errors with the short battery life of the N95 when

Bluetooth is activated on the phone The higher number of fall-events is due to the resending of these when

no acknowledgements were received.

Results – Fall AlgorithmResults – Fall AlgorithmDuring the trials a number of falls did occur and were

recorded onto the SD cards.Here is one of those falls

Results – Fall AlgorithmResults – Fall Algorithm

Results – The vestResults – The vestThe vests were worn for the full length of

the trial by all 10 elderly subjectsHowever feedback from the elderly subject

and the nursing staff indicated that the vests were not appreciated.

The elderly subjects felt the vest was uncomfortable and they disliked wearing it for eight hours each day.

The nursing staff felt that they too bulky and too intrusive if to be worn under clothes, along with the fall-sensor.

Results – The vestResults – The vestA number of improvements were

suggested:The vest should be made larger but

shorter so that the subjects have no difficulty in wearing and move about especially when using the toilet.

During the summer vest would not be appropriate for the high temperatures.

The vest should be made from more elastic material

Discussion\ConclusionDiscussion\ConclusionThrough incorporating the fall sensor into a vest that can

be worn by the elderly, it is considered that greater compliance with wearing and using a fall detection system can be achieved.

During the long term trials, 42 fall-alerts were recorded by the fall-sensor however only 9 were received at the care taker siteThus indicating that further development of the fall-

detection algorithm and the transmission protocol and method is required.

Discussion\ConclusionDiscussion\Conclusion

Also following feedback from the elderly subject’s it is clear that the vest were sufficient for a short term clinical trial.

However further development of the vest is required to make it more comfortable, breathable and easier to don and doff.

Further development of the system will include:more accurate fall-detection and fall-message transmission

algorithm,more comfortable method of attachment, lighter and smaller

sensor as well as, mobility monitoring and energy expenditure

measurement.

Questions

?Contact : [email protected]

See also: www.caalyx.eu

Thank you.