Doctor of Philosophy
School of Electrical, Computer and Telecommunications Engineering
Ameli, Sina, Muscular fatigue assessment based on inertial sensing and machine learning, Doctor of Philosophy thesis, School of Electrical, Computer and Telecommunications Engineering, University of Wollongong, 2017. https://ro.uow.edu.au/theses1/269
Prolonged sustained activities-induced muscular fatigue adversely affects physical performance. The most common side-effect of cancer and chemotherapy treatment is also muscular fatigue, which decreases the quality of life (QoL) and increases the chance of mortality. The degree of muscular fatigue is also a key factor in the design of a physical training and rehabilitation program for sports players. Conventional fatigue measurement methods are either question-based or invasive. The questionnaire-based methods are qualitative, subjective, inaccurate and prone to error. The current invasive methods cannot provide reliable information on the muscular fatigue of healthy subjects. On the other hand, the methods used in sports players and healthy subjects do not effectively assess fatigue in cancer patients. Overall, a reliable and objective muscular fatigue measurement tool which can be universally deployed across all subjects is not available.
In response to this gap in the literature, the development of an objective, non-invasive and universal muscular fatigue measurement tool that can reliably quantify the degree of muscular fatigue is investigated in this thesis. Fatigue is estimated based on gait and posture measured by MEMS wireless inertial sensors embedded in a motion capture system. A set of experimental procedures is designed to measure the fatigue caused by sustained activities in athletes and induced by chemotherapy in cancer patients. In the former, the physical performance of a subject doing 6 Minute-Walk (6MW) and Stair Climbing Test (SCT) is compared in pre- and post-interventions of a fatigue protocol. Whereas in the latter, the performance of a subject in 6MW and SCT is measured and compared in pre- and post-interventions of chemotherapy. Various classification methods such as Bayesian classifiers, adaptive soft computing, and time-series analysis are deployed to identify the effect of fatigue on motion characteristics and to quantify the degree of decline in ambulatory performance.
The performance of the developed algorithms is benchmarked against the physiological parameters measured during exercise-induced fatigue and are correlated with conventional methods used to assess fatigue in cancer patients. The analysis shows that the fatigue measured by the proposed algorithms increases the accuracy of conventional fatigue measurement methods and provides a reliable and objective alternative to the subjective methods used to measure fatigue in cancer patients.