New directions in recording and processing electro-mechanical signals from the human body

Voluntary contractions of muscles are electrochemically provoked by neural inputs which are initiated in the brain. In turn, these contractions generate forces that are mechanically transferred via tendons over joints causing joint torques and movement. The generation of muscle forces depends on a variety of factors. In fact, to model or predict human movement, knowledge on the activation level of muscles, their intrinsic mechanics, and fatigue are necessary parameters but hard to collect directly in-vivo. Therefore, this work aims to accelerate and improve direct, in-vivo quantification of human movement and suggests new methods for data collection and interpretation. Moreover, this thesis provides three open-source software packages and two datasets for future research and developments. The thesis is divided into three narrative streams which focus on ultrasound technology in different themes. The first part of this thesis deals with an assessment of the use of ultrasound in the biomechanical applications. Furthermore, it demonstrates a machine learning algorithm to accelerate muscle-tendon junction analyses in ultrasound images. We show that our proposed method provides human-like performance, requires only a fraction of manual labeling times (approx. 100 times faster) and transfers to previously unseen data. The second part focuses on the use of ultrafast ultrasound to study the mechanical response of muscles to their electrical activation. These observations exposed the electro-mechanical behaviour of sub-fascicular structures (approx. 280 - 430 sarcomeres in series). Furthermore, it provided the knowledge base for the development of an ultrasound transparent 1-channel electromyography system which is presented in the third part of this thesis. The developed interface allows parallel acquisition of 1-channel electromyography and ultrasound signals in the same muscle region without compromising signal quality.