Convergence of smartphone technology and algorithms that estimate physical activity for cardiac rehabilitation

Download files
Access & Terms of Use
open access
Copyright: Del Rosario, Michael
Altmetric
Abstract
Completing a cardiac rehabilitation program (CRP) following myocardial infarction has many health benefits, but CRP is not completed by all patients. This thesis investigated if completion rates could be improved by providing patients with a smartphone and telehealth equipment whilst enrolled in the CRP, and if there was a relationship between the amount of physical activity they completed during their enrolment, and the six minute walking distance. In order to estimate the amount of physical activity completed by participants whilst enrolled in a CRP, a model capable of recognising five human activities (postural transitions, stationary periods, walking on level ground as well as up and down stairs) by analysing the measurements from the smartphone’s internal sensors was developed using sensor data collected from both younger and older adults. This model (with an average total classification accuracy of 90.4% and average Cohen’s kappa of 0.83) was incorporated into an application which was installed on the smartphone of participants who were in the intervention arm of the study. Additional methods were also investigated that might enable the model to more accurately differentiate between standing, and sedentary periods in future. A computationally lightweight method – complementary attitude and heading reference system, was developed that estimated the attitude of a magnetic and inertial measurement unit (root mean square error in the pitch, roll and yaw angles of 1.84 degrees, 3.37 degrees, and 4.83 degrees, respectively). Incorporating this method into a new model for recognising human activities improved the model’s performance due to its use of an attitude invariant feature that calculated the angle between the average attitude during upright periods and the average attitude over the previous 2.5 seconds. This feature enabled the standing (class sensitivity 80%) and sedentary (class sensitivity 97%) classes to be separated, regardless of the smartphone’s attitude in the pants pocket. The results of a randomised controlled trial in which participants were recruited from a hospital-based CRP to receive the proposed adjunct or complete the standard CRP identified a significant difference in completion rates between treatment groups (88% vs 67%; p = 0.038) in favour of those randomised to the intervention group. This suggests that the telehealth adjunct increased the likelihood that participants would complete the program.
Persistent link to this record
Link to Publisher Version
Link to Open Access Version
Additional Link
Author(s)
Del Rosario, Michael
Supervisor(s)
Redmond, Stephen
Lovell, Nigel
Creator(s)
Editor(s)
Translator(s)
Curator(s)
Designer(s)
Arranger(s)
Composer(s)
Recordist(s)
Conference Proceedings Editor(s)
Other Contributor(s)
Corporate/Industry Contributor(s)
Publication Year
2017
Resource Type
Thesis
Degree Type
PhD Doctorate
UNSW Faculty
Files
download public version.pdf 14.99 MB Adobe Portable Document Format
Related dataset(s)