Lower back muscle activity and fatigue during activities of daily living using a novel wearable device and machine learning
File(s)
Author(s)
Terracina Barcas, Dan-Emanuel
Type
Thesis
Abstract
Low back pain (LBP) is one of the major causes of disability in the modern world. People affected tend to experience an altered muscle activity and fatigue from the trunk muscles. In this thesis, the aim is to study trunk muscle activity and fatigue from a physiological and computational perspective in laboratory and real-world conditions. In order to achieve this goal, the thesis combines experiments mimicking real-world conditions in laboratory and real-world environments as well as machine learning, adaptive signal processing and control theory to study, evaluate and predict the levels of muscle activity and fatigue. The work begins with an explanation of the complexity of assessing muscle activity and fatigue using surface electromyography and highlighting the need for a wearable technology to perform long term monitoring of the lower back muscles during activities of daily living (ADL) to deepen the understanding of the functioning of these muscles, and potentially LBP. Then, it continues with a laboratory experiment simulating ADL in a controlled environment and showing how ADL affect the ability to exert the same levels of force as shown by the decrease in the maximum voluntary contraction of the lower back muscles following a 30-minute walk. Following this and with the idea of preventing rather than reacting to muscle fatigue, novel algorithms aiming at forecasting, classifying, and modelling the levels of muscle fatigue and activity 25 seconds ahead of time are presented. The superior performance of the dilated convolutional neural network is shown, outperforming four adaptive signal processing algorithms by 30.0% to 82.8% according to the feature evaluated. The benefit of these algorithms is that they can be implemented in a wearable platform given their little computational complexity. Finally, the final part of this work intends to answer one of the key questions of this thesis: how do trunk muscle activity and fatigue vary in real-world conditions? For that, a wearable device that enables long term monitoring of trunk muscle activity during ADL in a real-world environment is presented. Bounded with dry electrodes of 10 mm and 15 mm with a superior signal-to-noise ratio, this novel platform overcomes the barrier of current systems that are limited to laboratory conditions, therefore hindering the possibility of long term monitoring outside of this environment. This final real-world study shows that LBP subjects have a different longissimus activity during sitting, lying, standing and walking. One limitation of the previous system is its limited spatial resolution, which hinders the understanding of how the muscle redundancy present in the lower back influences the fatigue and contraction patterns. Hence, in order to close this gap, this work presents a second version of the platform with a custom application-specific integrated circuit (ASIC) extracting in real time two of the most common muscle fatigue and activity features: the root mean square and the zero-crossing. Future work will intend to merge the new ASIC with the custom 32-channel array of electrodes mapping the entire lower back and therefore unveiling the effects of trunk muscle redundancy, as well as embedding the muscle fatigue forecasting algorithms presented in this thesis. The latter could provide a deeper understanding of the contraction patterns during ADL.
Version
Open Access
Date Issued
2021-07
Date Awarded
2021-11
Copyright Statement
Creative Commons Attribution NonCommercial Licence
License URL
Advisor
Strutton, Paul
Georgiou, Pantelakis
Sponsor
Engineering and Physical Sciences Research Council
Publisher Department
Department of Surgery & Cancer
Publisher Institution
Imperial College London
Qualification Level
Doctoral
Qualification Name
Doctor of Philosophy (PhD)