QUANTIFYING ENERGY EXPENDITURE OF WHEELCHAIR-BASED PHYSICAL ACTIVITIES IN FREE-LIVING ENVIRONMENTS SHIVAYOGI HIREMATH UNIVERSITY OF PITTSBURGH
Jun 19, 2015
QUANTIFYING ENERGY EXPENDITURE OF
WHEELCHAIR-BASED PHYSICAL ACTIVITIESIN FREE-LIVING ENVIRONMENTS
SHIVAYOGI HIREMATH
UNIVERSITY OF PITTSBURGH
Quantifying Energy Expenditure of Wheelchair-based Physical Activities in Free-living Environments
Shivayogi Hiremath, Stephen Intille Rory Cooper, Dan Ding Wireless Health 2014 Bethesda, MD, USA
Introduction • Lack of regular physical activity (PA) is a major
public health concern
• Regular PA is crucial in wheelchair users as it is associated with
– Aerobic Capacity and Flexibility – Muscular strength and Endurance – Improved psychological well-being – Risk of cardiovascular disease and other chronic
conditions • Regular PA levels in wheelchair users are low
– Environmental barriers – Lack of accessible equipment – Physiological changes
© Shivayogi Hiremath 2010-14
2
Introduction • Research has shown that it is possible for
wheelchair users to attain PA levels recommended by the ACSM1
– Wheelchair basketball, tennis, Handcycling
• Self-monitoring of diet, PA and body weight can assist in maintaining a healthy lifestyle
• Availability of an activity monitor will allow wheelchair users to achieve optimal regular PA leading to a healthy and active lifestyle
3 © Shivayogi Hiremath 2010-14
Background • Sensor-based PA monitors have been used
to track and quantify PAs among wheelchairs users 2-6
– None of these monitors can capture and provide real-time feedback to the user
Approach • Real-time feedback enabled by
advancements in m-Health • Meaningful feedback to wheelchair users
© Shivayogi Hiremath 2010-14 4
5 © Shivayogi Hiremath 2010-14
Physical Activity Monitoring System
Wocket7
G-WRM8
PAMS
Testing of PAMS in MWUs • 45 MWUs with SCI
– Laboratory (n=25) & NVWG 2012 (n=20) – Home (n=20)
• Activities Performed – Resting – Arm-Ergometry – Darts, Basketball – Deskwork, Watching TV – Folding Clothes, Laundry – Food Preparation, Eating Simulation – Propulsion: Carpet, Tile, Ramp, Home – Resistance: Band, Dumbbell, Handgrip – Making Bed, Floor Sweeping – Cleaning Room, Vacuuming – Wheelchair Pushups
© Shivayogi Hiremath 2010-14
6
Model Development • Two step classification algorithm to detect
wheelchair based PAs • Activity-specific energy expenditure
estimation models – Resting – Arm-ergometry – Household activities – PAs that might involve wheelchair movement – Wheelchair propulsion – Caretaker pushing – Basketball
© Shivayogi Hiremath 2010-14
7
Results
-1
-0.5
0
0.5
1
1.5
2
-3
-2
-1
0
1
2
3
4
5
6
1 1 1 1 1 2 2 2 2 3 3 3 3 4 4 4 4 5 5 5 5 6 6 6 6 7 7 7 7 8 8 8 8 9 9 9 9 10 10 10 10 11 11 11 11 11
Dis
tan
ce T
rave
lled
in m
/10
s
Acc
eler
atio
n in
m/s
2
Activity Trials
Mean Acceleration and Distance Travelled WocketArmDistTravelled
Resting Arm-Ergometry Darts Deskwork Folding Propulsion Invest Resist- Clothes Carpet Moderate Slow Pushing ance
© Shivayogi Hiremath 2010-14
8
Results Demographic Characteristics
Age 41.0 ± 12.6 years
Gender 39 males; 6 females
Weight 78.1 ± 18.1 kg
Height 1.8 ± 0.1 m
SCI Level C5 to L5
Completeness 22 (Complete); 23(Incomplete)
Wheelchair use 12.6 ± 8.6 years
Smokers 14
PA participation 36 regular; 5 occasional; 4 no regular PA
9 © Shivayogi Hiremath 2010-14
Results
10
00.5
11.5
22.5
33.5
44.5
5
EE
Energy Expenditure for Various Activities
EE in kcal/min
© Shivayogi Hiremath 2010-14
Results: Classification 80-20CV
© Shivayogi Hiremath 2010-14 11
Training Accuracy Testing Accuracy Models
0.9356 0.9801 SVM
• First level classification
M/NM Training Accuracy Testing Accuracy Model
NM 0.8678 0.8495 J48
M 0.9557 0.9340 SVM
• Second level classification
• Classification into seven PAs showed that PAMS had an overall accuracy of 89.3%
Results: Confusion Matrix
© Shivayogi Hiremath 2010-13
True\Predict R AE NM Prop Push Bask MM
Resting 22 0 24 0 0 0 1
AE 0 85 6 0 0 0 1
PAs NM 3 2 135 0 0 0 0
Propulsion 0 0 0 134 1 0 0
Pushing 0 0 0 2 44 0 1
Basketball 0 0 1 5 1 17 6
MM 0 0 0 0 0 0 12 12
Results: Mean Signed Error
13
-50-40-30-20-10
01020304050
Per
cen
tage
Err
or (
%)
Wheelchair based Physical Activities
Plot of Mean Signed Error for PAMS
PAMS
© Shivayogi Hiremath 2010-14
EE estimation error post-classification showed that the overall EE error for PAMS was -9.8% (0.1%).
Results: Mean Absolute Error
14
05
101520253035404550
Per
cen
tage
Err
or (
%)
Wheelchair based Physical Activities
Plot of Mean Absolute Error for PAMS
PAMS
© Shivayogi Hiremath 2010-14
Results: Metabolic Equivalent Tasks
15
METs based Intensity Actual mins Estimated mins K4b2 PAMS
Light Intensity 387 406 Moderate Intensity 116 97 High Intensity 5 3
© Shivayogi Hiremath 2010-14
Discussion
• Frequency based features were picked for classification algorithms for certain PAs
• Majority of the regression equations chose demographic characteristics to estimate EE
• Multi-modal information from PAMS can be utilized to quantify wheelchair-based activities in laboratory and community
• Limitations –Active population –Movement based variables
© Shivayogi Hiremath 2010-14
16
Future Work
© Shivayogi Hiremath 2010-14
• Evaluation of PAMS in community settings
• Evaluation of PAMS in community along with DLW
• Expand to other population • Use social networks such as Facebook
– Combine self-monitoring with social support
17
1. American College of Sports Medicine, Mitchell H. Whaley, Peter H. Brubaker, Robert Michael Otto, Lawrence E. Armstrong ACSM's guidelines for exercise testing and prescription, 7, illustrated ed.: Lippincott Williams & Wilkins, 2005.
2. C. A. Warms and B. L. Belza, "Actigraphy as a measure of physical activity for wheelchair users with spinal cord injury," Nursing Research, vol. 53, pp. 136-43, 2004.
3. S. E. Sonenblum, S. Sprigle, J. Caspall, and R. Lopez, "Validation of an accelerometer-based method to measure the use of manual wheelchairs," Medical Engineering and Physics, vol. 34, pp. 781-86, 2012.
4. M. L. Tolerico, D. Ding, R. A. Cooper, D. M. Spaeth, S. G. Fitzgerald, R. Cooper, et al., "Assessing mobility characteristics and activity levels of manual wheelchair users," Journal of Rehabilitation R and D, vol. 44, pp. 561-72, 2007.
5. M. Lee, W. Zhu, B. Hedrick, and B. Fernhall, "Estimating MET values using the ratio of HR for persons with paraplegia," Med Sci Sports Exerc, vol. 42, pp. 985-90, 2010.
6. E. H. Coulter, P. M. Dall, L. Rochester, J. P. Hasler, and M. H. Granat, "Development and validation of a physical activity monitor for use on a wheelchair," Journal of Spinal Cord, vol. 49, pp. 445-450, 2011.
7. S. S. Intille, F. Albinali, S. Mota, B. Kuris, P. Botana, and W. L. Haskell, "Design of a Wearable Physical Activity Monitoring System using Mobile Phones and Accelerometers," in Proceedings of Annual International Conference of the IEEE Engineering in Medicine and Biology Society. , Boston, MA, 2011, pp. 3636-39.
8. S. V. Hiremath, D. Ding, and R. A. Cooper, "Development and evaluation of a gyroscope based wheel rotation monitor for manual wheelchair users.," Spinal Cord Medicine, vol. 36, pp. 347-356, 2013.
© Shivayogi Hiremath 2010-14
18
References
Acknowledgements
• Switzer Research Fellowship (H133Fl10032), NIDRR, Department of Education
• Department of Defense (W81XWH-10-1-0816) • RERC on Interactive Exercise Technologies
and Exercise Physiology for Persons with Disabilities (H133E070029), NIDRR
• VA Center of Excellence for Wheelchairs and Associated Rehabilitation Engineering (B3142C)
© Shivayogi Hiremath 2010-14 19
www.wirelesshealth2014.org