Multi-scale Analysis Of Surface EMG Signal During Muscle Fatiguing Process In Children With Cerebral Palsy

Muscle fatigue is a physiological phenomenon of the muscle’s inability to perform the original activities due to the decline in the output force of the muscle after a sustained contraction. The appearance of muscle fatigue has, to some extent, been a limitation in rehabilitation training and functional activities of daily life for children diagnosed with cerebral palsy (CP).Therefore, investigation into muscle fatigue characteristics of children with CP is very important for their clinical diagnosis and motor function evaluation. Surface electromyography (sEMG) has been commonly used in the assessment of muscle fatigue, by reflecting a variety of central and peripheral changes during muscle fatiguing processes. The multi-scale analysis of sEMG is able to reflect the states and levels of activities of the neuromuscular system at different scales. In this thesis, multi-scale analysis of sEMG recorded during muscle fatiguing process was performed for children with cerebral palsy. Changes in underlying motor control mechanisms as a result of CP was preliminarily investigated by revealing differences in the experimental results between children with CP and typically development (TD).A novel testing protocol for examining upper-limb muscle fatigue during dynamic task performance was deigned, which was suitable for home rehabilitation for children with CP. A total of 34 subjects participated in the experiment, including 18 age-matched TD children and 16 CP children. By measuring the maximum lifting force (MLF) prior to the experiment, three different loads, namely 0%,30% and 60% of the MLF, were obtained for each subject. During the experiment, each subject was asked to perform cyclic upper-limb lifting tasks with each of the three loads, while the sEMG signals were recorded and stored for further analysis.The original sEMG signals were decomposed into 7 intrinsic mode functions (IMFs) via empirical mode decomposition (EMD). After determining the onset and offset of a sEMG segment corresponding to muscle contraction during upper-limb lifting in each cycle, three parameters, namely the mean power frequency (MPF), root mean square (RMS) and sample entropy (SampEn) were computed for each segment of the original sEMG and resultant 7 IMFs, respectively. The differences in the experimental results between TD and CP children were also evaluated. The primary findings were listed as follows:(1) During the fatiguing progress, MPF of the original sEMG signals from the TD children showed a declining trend, RMS showed an increasing trend, and SampEn presented a declining trend. These trends became more evident with the increase of loads. However, only SampEn of the original sEMG signals from the CP children showed a declining trend, which became evident with the increase of loads as well. (2) After the EMD decomposition of the original sEMG signals, the three parameters of the lower order IMFs from the TD children showed similar changing trend with those of the original signals during the fatiguing process. For the CP children, the MPF of IMF 1 showed an evident declining trend as compared with the original signal. In addition, the SampEn of the IMF1 showed a corresponding changing trend with that of the original signals. However, the three parameters of other IMFs showed no significant changes with the increase of muscle fatigue level.After the entire sEMG signals were evenly segmented into three signal windows with an equal length (denoted as W1, W2 and W3, representing different degrees of muscle fatigue), multi-scale complexity of each signal window was evaluated by performing two multi-scale entropy (MSE) methods, namely the standard MSE method based on the "coarse-grained" approach and the EMD-enhanced MSE method, respectively. The resultant MSE curve were then obtained for each signal window. The effects of the muscle fatigue and load on the multi-scale complexity of sEMG were investigated through comparing differences in MSE curves between signal windows and under three different loads (0%,30%,60% MLF) respectively. Moreover, the differences in MSE results between TD and CP children were compared as well. The experimental findings can be summarized in the following two aspects:(1) As the level of muscle fatigue (from W1 to W3) was increased, the declining trend of MSE curves from the TD children was reduced with statistical significance. By contrast, there was not any significant changes in the declining trend of the MSE curve for the children with CP. (2) The declining trend of MSE curves from the TD children became more significant with the load increase, whereas such observation was not found for the CP children.On all accounts, the experimental findings about the muscle fatigue may be attributed to many factors such as decreased muscle fiber conduction velocity and increased motor unit discharge synchronization. The designed testing protocol along with the proposed multi-scale analysis of sEMG produced several useful indicators of muscle fatigue. On this basis, the revealed differences in experimental results between the CP and TD children reflected the complex changes at work as a result of CP such as selective loss of larger motor units and the increased level of motor unit discharge synchronization. This study helps to reveal neuropathological mechanism of children with CP, and also provides important guidelines for the clinical diagnosis and motor functional evaluation of children with CP.