The development of Muscle Fatigue Prediction model from muscle torque and contraction data / Keshasni Earichappan

• Main Author: Keshasni , Earichappan
• Format: Thesis
• Published: 2023
• Subjects: RA Public aspects of medicine; TA Engineering (General). Civil engineering (General)
• Online Access: http://studentsrepo.um.edu.my/15539/1/Keshasni.pdf; http://studentsrepo.um.edu.my/15539/2/Keshasni_Earichappan.pdf; http://studentsrepo.um.edu.my/15539/

The growing need for a reliable method for assessing muscle function continues to facilitate further research into other alternatives. Therefore, a novel MC sensor is an alternative tool to study muscle function. This study aims to develop a prediction model to predict muscle fatigue levels (100%-75%, 75%-50%, 50%-25%, 25%-0%) from muscle contraction and torque using Levenberg Marquardt algorithm. This study starts with the collection of data on muscle contraction(mV) and torque (Nm) based on six healthy subjects doing isokinetic knee extension. Muscle torque was measured using mechanomyography and Biodex dynamometer sensors. Then, the data obtained were first normalized, and the peak values of both signals were analyzed and categorized into four different muscle fatigue levels. The correlation between MC and torque for each subject were identified every 5% for MC-duration vs torque and MC-general vs torque. The MC graph is then overlayed with the torque graph to find the relation between them starting from the highest peak of the signal. The average MC and torque data were set as input parameters and binary formed fatigue levels were set as output parameters. The optimum model has been developed using hidden layers of 10 neurons, threshold 0.5 and kfold of 5. Results show significant drop in MC sensor voltage and torque value concurrently with time, with MC-duration having better correlation with torque (r is equal to 0.99465) compared to whole MC signal vs torque (r is equal to 0.32807). The dynamometer isokinetic knee torque and MC sensor data had a significant linear connection, which indicated that the MC sensor could detect different levels of muscular contraction and a fatiguing contraction in persistent voluntary contraction in healthy individuals. The results show that the peak time (s) of MC matches with the torque throughout the four muscle fatigue stages. It can be observed that the muscle contraction signal and muscle torque are corresponding as the time and fatiguing percentage increase especially during the fatigued muscle stage. It can also be observed that the muscle contraction signal peaks coincide with torque’s peaks throughout the fatigue percentages. MC signal significantly captured the muscle contraction in matching correlation duration with torque signal. The developed prediction model predicted muscle fatigue with the accuracy of 67.37% and proved there is a strong matching between muscle torque fatigue and muscle contraction fatigue with sensitivity of 80.55%. This study illustrated the potential connection that might exist between muscle fatigue with muscle contraction and muscle torque in healthy persons.