Characterizing Muscle Fatigue with Topological Data Analysis
Author:
Allyson Clarke ’24Co-Authors:
Benjamin B. Wheatley, Chulhyun AhnFaculty Mentor(s):
Benjamin Wheatley, Mechanical EngineeringFunding Source:
PURAbstract
Half of stroke survivors require long-term rehabilitative care, which is often complicated by a high degree of deficit variability. One debilitating effect of a stroke is muscle fatigue caused by muscle weakness. Identifying fatigued muscles can allow for targeted treatment plans and quicker rehabilitation; however, such a strategy has been hampered by difficulties in accurately characterizing fatigue.
Various linear and nonlinear signal processing methods have been used to characterize muscle fatigue from surface electromyography (sEMG) data, but they only characterize the frequency domain. Efforts to capitalize on the topological properties of sEMG, which include frequency and amplitude information, are in their incipient stages.
My research explores using topological data analysis as a robust measure of muscle fatigue. I constructed a custom device that stabilizes the hand and processed the data using a frequency-based Fourier transform. Leveraging techniques from Chutani, my team and I extracted a topological property (number of simplices) from the raw sEMG data. I compared the results of the two analyses and determined that the number of simplices is a more accurate indicator of fatigue than the traditional process, laying the framework for a new method of fatigue detection that could have meaningful implications for rehabilitation and sports sciences.