Human movement results from contractions of muscles spanning joints. Within muscles are fibres which contract in groups in response to electrical impulses from motoneurons. Together, the motoneuron and its innervated fibres make up the motor unit (MU). The speed and force of muscular contractions are modulated by the number of recruited MUs and their firing rate. The ensemble of all MU discharge times constitutes the neural drive to the muscle. Thus, to be able to study MUs is to be able to study the fundamentals of human movement. Further, to be able to interpret intended movement from neural drive presents a powerful tool for human-machine interfacing.
Whilst the surface electromyogram (sEMG) has classically been utilised for this, it suffers from low spatial resolution, crosstalk between muscles, and small detection volume. In this thesis, ultrafast ultrasound imaging, capable of capturing thousands of frames per second, is proposed as an alternative technique. If the contraction of the muscle fibres can be captured using ultrasound, the precise activation time of the innervating motoneurons may be deduced. Here, ultrafast ultrasound has clear advantages over sEMG owing to its high spatial resolution and specificity, high temporal resolution, and high penetration in tissue.
The thesis opens by establishing the use of ultrafast ultrasound for the kinematic study of MU motion during voluntary contractions. Next, it is shown that while previously proposed source separation methods can extract MU locations from ultrasound images, they cannot accurately extract motoneuron discharge times, and therefore cannot serve as neural interfaces. Hence, the thesis culminates in the proposal and validation of a novel model for source separation which accurately estimates motoneuron discharge times.
Overall, the work in this thesis has enabled extraction of neural drive from an ultrafast ultrasound image series, paving the way for a new generation of human-machine interfaces.