Development of a machine learning-based lower-limb exercise training system for individuals with knee pain and empirical evaluation of its usability, acceptance, and health effects

Abstract

Healthcare technologies have been increasingly developed to support the delivery of healthcare services and promote health management. However, their development and implementation do not always meet the needs of end-users, which can pose barriers to their usability, acceptance, and health effectiveness. This thesis proposed a machine learning-based system to facilitate lower-limb exercise training in middle-aged and older individuals with knee pain. The system offered three key features: video-based exercise demonstrations, real-time movement feedback, and performance and progress tracking. Adopting a user-centered approach, design and development of the system were conducted over five phases. Usability and user acceptance of the system were iteratively investigated. The preliminary insights into its effects on knee health were also provided. Phase I introduced the initial design of the system. A pilot study was conducted to capture the initial views among eight participants. They used the system to perform exercises in the study laboratory for three consecutive days. Their perceptions and suggestions gave directions to future development of the system. In Phase II, a paper-based prototype system was developed. A cross-sectional survey was conducted to examine the perceptions among the target population, that is, middle-aged and older adults with knee pain (n = 94). Most participants (> 75%) expressed positive opinions about the system’s perceived effects, perceived ease of use, and features; they also showed intention to use the system if given an opportunity. The findings highlighted the need to develop a computer-based system and further investigate its usability and user acceptance. In Phase III, a computer-based prototype system was developed. Human factors practitioners identified four design deficiencies through a heuristic evaluation, regarding recognition and recovery of errors, navigation, auditory perception, and help documentation. In addition, an end-user testing was conducted. Ten individuals with knee pain used the prototype system to complete five tasks in the laboratory. In general, the prototype system was perceived as usable and acceptable. It is worthwhile to further develop and evaluate a fully functional system. In Phase IV, the system was implemented among 60 individuals with knee pain to assess its usability and acceptance. Preliminary evaluation of the system for knee health was also conducted. Participants came to the laboratory to use the system for three weeks to undergo six sessions of lower-limb exercise training. The system was found to be usable and acceptable. The experimental findings also demonstrated the short-term benefits of the system in terms of physical function measured by the Western Ontario and McMaster Universities Osteoarthritis Index, quality of life, and exercise engagement. Finally, in Phase V, the system design was improved and finalized. By addressing the usability issues identified in the previous phases, the system’s usability and ease of use were enhanced. This study contributed to design, development, and empirical evaluation of advanced healthcare digitalization. The use of user-centered design helped to create an exercise training system that was user-friendly and easy to accept. This study provided preliminary evidence of its short-term benefits, while further investigations are warranted to assess the long-term effectiveness in managing knee pain.