Do low back pain patients show an abnormal activity pattern?

Remote Monitoring &
Remotely Supervised Training
Hermie J. Hermens, PhD
Roessingh Research &
University of Twente
Trends in health care
• Rising demand for care
– Increase of number of elderly people
– Increase number of people with chronic disorders
• Rand Coop predicts in US
• 2010: 141 million people with chronic illness
• 2030: 171 million people with chronic illness
– Chronic Care now over 50% of total costs and
Costs of Health Care (% BNP)
• Source: OECD (2002)
Tsjechie 5.7 % BNP
England: 7.7 % BNP
Netherlands: 9.1% BNP
Germany 10.5 % BNP
Suisse 11.2 % BNP
United States 14.6 % BNP
Quote Martin van Rijn (NL): with unchanged policies,
health care will cost us 12% BNP and 20% of working
people will have to work in healthcare
Results of these trends in health care
• Rising demand for chronic care
– With limited budget and people
– Will force a higher productivity without losing our
high quality
• Changes in the customer will require
– Independent living as long as possible
– Individually tailored solutions
All these changes will require a change in our
approach to health care
Need to change our approach
Present focus
From Philips
Need to change our approach
Required extensions
From Philips
Trends in Technology
• Sensing technology
High quality ambulatory acquisition possible
Smaller, low power, wireless connection
User friendly (integrated in textiles)
PDA’s get smarter and more powerful
• Information & Communication Technology
– Broadband connection available and cheap
– Special data transport platforms available
– Centralised Electronic Health record available
Creating new opportunities
Combining Biomedical Engineering with Information
and Communication Technology creates a new
area of research and relevant clinical services:
Remote monitoring
Remotely supervised training
Enabling monitoring and treatment of subjects
anywhere, anytime and intervene when needed
Remote monitoring
Guarding the health condition of a subject
by measuring and interpretation of vital
biosignals, without interference of his
daily activities but able to react when
Remotely supervised training & Treatment
Enabling the subject to train at his time
and place, providing him the same
quality of feedback/assurance as in
the intramural situation
= Monitoring + dedicated feedback
The potential benefits
– Less intramural care (costs)
– More freedom for the patient
– Peace of mind
Remotely supervised treatment
– High intensity of training possible (more = better)
– Training in natural environment translates better to ADL
situations (more effective training)
– Patient himself responsible for results
– Clinician can ‘treat’ several patients at the same time
The challenges
Are we able to make this technologically feasible ?
Will this result this in a same quality of treatment ?
Will this be accepted by health care providers ?
Will this be accepted by the patients ?
Summarising experiences of the past ten years by:
Roessingh Research & Rehabilitation Centre &
University of Twente
Many partners (Lucent, Philips, Atos Origin, ..)
Ongoing research focused, on ambulatory
monitoring and treatment
Remote monitoring
remotely supervised training
General architecture RMT systems
Decision support
Care & Coaching
(Hermens, 2008)
An example: The Mobihealth system
 Developed in European projects
(Mobihealth and HS24, Awareness)
 Supports mobile data transport
 Supports various networks
 Data encryption of biosignals
 Access
 User Identification by password
 Device authentication by pin
 Tested in many clinical pilot studies
secure wireless transmission
patient data
 See
Service Centre
hybrid data communication infrastructure
Sensing; general demands
 Sensors
 Sensors should be wearable, comfortable, forgettable
 Autonomic placement feasible
 Processing
 Continuous sensing required, independent of place, time
 “Real-time” processing and feature extraction required
Sensor development in Twente
Capacitive EXG
EMG garment
Activity monitoring
3-D Force shoe
Full body movements
Automatic feature extraction in EMG
E.g. spasticity
(Detect when and how
often muscle active)
AGLR to detect
changes in variance
Post-processing based
on physiol. properties
Looking for new features: physical condition
Important variable in
chronic diseases
But requires max effort
Can this be estimated in
non max conditions
Have people do various
activity , while measuring
ECG and activity
ECG and activity
Estimation of physical condition from
ambulatory measurements
First results :Modeling predicts good correlations with Astrand
cycle ergometer test and modified Bruce treadmill test
The Body Area Network (BAN)
 Often more then on body sensor is required (sensor fusion):
 To enable more robust features (e.g. Movements)
 To enable different features (e.g. Physical condition)
 And actuators for feedback purposes
Extra BAN
Intra BAN
BAN boundary
So, we need to connect multiple sensors and actuators to a central
point to enable synchronised data collection
Several approaches to create a BAN
One amplifier/AD converter (classic)
Bussystem (e.g. Xsens)
Multiple bluetooth connections
Upcoming: Wireless sensor networks
Applications = Services
What kind of services should we develop?
Considering that chronic diseases develop slowly but sudden
events might happen
Monitoring services should aim at:
 Monitoring sudden adverse events
 Detecting slow changes over time
 Detecting changes in patterns
 Treatment services should aim at:
 Providing feedback to the person, so he can change his
“negative” behaviour and
 To the health care professionals for consultation purposes
support and to enable interventions
Case 1
Chronic low back pain
A Tele-treatment of chronic pain patients
• 80% all people ever have low back
• About 90% recovers, 10% becomes
• Over 80% no clear damage
• Lot of medical shopping
• High costs (5 BE, 1995)
• Present treatments not very effective
(35% for multidisc. Programs)
• All models do predict a change in
activities as part of the chronification
Do low back pain patients show an abnormal
activity pattern over the day?
• Chronic pain patients (n=29) and
asymptomatic controls (n=20)
• Wore MT9 inertial 3-D motion sensor to
measure the activity level during 7
consecutive days.
• Fill in questionnaires to assess the
activity level subjectively.
Van Weering et al 2007
Do low back pain patients show an
abnormal activity pattern?
mean acceleration
• Overall activity level not different between patients and controls
• Activity pattern Patients unbalanced: significant higher in the
morning and lower in the evening
An idea for a new treatment concept
• Starting from: LBP patients have a
dysbalanced activity pattern during the day
• Assuming that such dysbalance in activity
is an important component in the
chronification process in low back pain
• Conceptual idea: Normalising this activity
pattern might reverse the chronification!
• Realise this by providing continuous
personalised feedback on the activity pattern
Personalised context aware feedback
After making breakfast
for the kids and while
doing the dishes, the
system detects that
Cinderella has been
too active for a period
of time. She receives
General feedback:
Personalized feedback
That she needs to rest for some
That she should have some tea.
Context-aware feedback (time,
weather, presence)
Drink a cup of tea in the backyard
and enjoy the sun.
The M-health service platform
context data
Back End
Body area network
Sensors &
context data
Mobile Base Unit
reasoning and storage
Continuous context aware feedback
Professional feedback
Medical display
Present status
Activity sensing implemented on PDA
Personalised messaging implemented
Context aware feedback not yet
Clinical trial recently started
First responses positive
Input pain level
Personalised feedback
Feedback of performance
Case 3.
Fully ambulatory training of
neck/shoulder pain
Chronic neck/shoulder pain
 Chronic pain in neck/shoulder with no
clear cause of physical overloading
 Often associated with computer work
 Cinderella theory:
 Lack of relaxation results in
overloading specific muscle parts
 Overloading results in pain
 Pain results in changes in posture
and more overloading
Solution: warn the subject in case of insufficient relaxation, so
subject is able to learn and adapt posture and muscle
Starting points for the system design
• Assess muscle relaxation by surface EMG measurement
and processing
• Provide private feedback when there is insufficient
• Enable an intense treatment outside the hospital!
– during normal activities: fully ambulant
– Non-obtrusive
– Support independent
EMG sensing garment for neck/shoulder
Summarizing our experiences
In about 50 patients:
• Unstable signals during first five
minutes, then good signal for over
24 hours
• Requires initial individual fitting,
then reproducible signals
• Independent donning and doffing
• Not interfering with activities of
daily living
EMG Processing for Feedback
Calculate relative relaxation
time (RRT; (resampling 125
ms; moving window 1 m.)
If RRT<20%, warn
subject with vibration
Myofeedback in practice
 Able to improve muscle relaxation (international RCT)
 Able to decrease pain complaints
But often intensive supervision required:
 Discuss experiences and results
 Troubleshooting in first week(s)
So, could this treatment be improved using ICT ?
Voerman et al 2006; Huis intVeld 2007)
The service system to enable remote
Sensors &
Web based
Database with signals
And subject data
Autonomous Feedback
Mobile Base Unit
Consultation & Feedback
From health care professional
Remotely supervised Myofeedback for
treatment of neck/shoulder pain
Initial Questions:
- Is it technically feasible to
monitor muscle activity during 8
hours per day?
- Is it accepted by patients and
care givers?
- Are care givers able to provide
advice not seeing the patient?
- Is it effective?
Huis intVeld et al. 2007
Results Remotely supervised Myofeedback for
treatment of neck/shoulder pain
- Inital study in 10 patients
- Often failure of wireless connections
- Enough data was received at backend
- Remote consultation feasible
- Confidence in treatment both by
patients and clinicians
- Clinically at least as effective as nonremote myofeedback treatment
Now entered market validation study in
3 countries (eTEN project Myotel)
Huis intVeld et al. 2007
The next step: show large scale feasibility
 Show effectiveness and efficiency (Market
validation ) in 3 countries (eTEN Myotel)
 Development business plan
 Development of Decision support system
Development of a CDSS
 To assist the clinician and patient in optimising advices during
consultation session
 Streaming data: 2 signals (RMS, RRT) of two muscles
Filtered, re-sampled at 4 Hz and stored in database
Together with activity diary and pain scores
Direction of solution
 Using Bayesian network to
 Detect technical failures
 Relating specific activities to pain
 Relate specific moments to related pain
Implement this in an agent platform
Case 2b
Monitoring of low back muscle
A similar feedback treatment for low back pain ?
• Literature shows inconsistent data on
the muscle activation of the low back
• Indications of both inactivity and
hyperactivity patterns were found
• So, what EMG patterns can be found
and do they differ from normal
Ambulatory measurement of low back muscles
• Special garment to measure the EMG
• Utilising dry electrodes in a flexible
system to enable stable contact during
activities of daily living
• Able to don/doff independently without
affecting the EMG signals ?
• Differences in muscle activity pattern
between patients and controls
• How, what, when to feedback ?
(de Nooy et al, 2008; patent pending)
Example EMG patterns low back muscles
Accumulated EMG activity in the evening:
Left healthy subject showing phasic patterns and
Right a patient showing rather continuous low level activation
Present status
• Garment can be worn during at least 8
hours without pain/serious discomfort
• Sensitivity misplacement: low in
longitudinal direction, high in lateral
• So far: data of 10 patients and 9 healthy
subjects during 7 days
• Differences in patterns apparent
• Feedback strategy : avoid constant
activity ?
• First implementation carried out
Present and future
Remote Monitoring and
Treatment (RMT)
Present status Remote Monitoring & Treatment
 75% applications fail in valorisation phase (Berg)
Technology not mature
Feedback primitive, not encouraging
 Health care organisations not ready yet
 Upscaling not yet possible (Decision support missing)
 Clinical evidence limited to pilot studies
Present status Remote Monitoring & Treatment
 Technology (sensors, ICT) rapidly developing
 Health Insurance companies get interested
 On agenda of EC, national research agendas
 RMT market 5.6 BD, growing with 70% per year (Liebert)
 2-5 years for mainstream adoption (Gartner)
[Gartner 2006]
Focus/trends of RMT in the next years:
intelligent autonomic personal health systems
- Biomedical Technology: creating comfortable,
accurate and robust sensing systems
-ICT: creating scalable, dependable, autonomic,
intelligent systems
- Integration with Ambient Assisted living
- Creation of virtual communities for support
and motivation
-Start with drugs delivery systems with
implanted sensors and remote support of
complex supportive systems
Remote Monitoring & Training
Very challenging area,
Requiring strong interdisciplinary
Still in its infancy period
But with a great promise
Thank you for your
Many Thanks to:
•Colleagues from Roessingh Research
•Colleagues from University of Twente
•Colleagues from Roessingh Rehabilitation centre