Search

Muscular activity is traditionally measured with surface electromyography (sEMG), which mostly reflects the activity of the superficial muscles. Ultrasound imaging (USI) is capable of investigating deeper structures and detecting very small contractions. The advancements in image processing led to development of automated methods based on tracking of features in US frames to analyse muscle properties. However, the analysis is typically very computationally expensive and can often only be performed offline. Real-time ultrasound is frequently used as a feedback tool by showing US videos to the participants, leading to improved control over the muscles. It does, however, not provide any quantitative information on the behaviour of muscle structures. Brain computer interface (BCI) experimental paradigms often rely on motor imagination (MI). Among other factors, it relies on proprioception, which is the ability to sense movements within joints and experience sensation from the muscles, therefore muscular activity needs to be carefully controlled, as the paradigms used on healthy populations might not replicate this process in people with impaired sensory-motor processing. The aim of this project is to combine information obtained from USI and EEG to provide a more detailed assessment of muscle behaviour and associated cortical activity during real and imagined movements. Furthermore, the research aims at providing biofeedback of muscle activity in semi-real time based on USI analysis for rehabilitation and muscle training. The first study presented in this thesis involved simultaneous recording of cortical activity electroencephalography (EEG), muscle activity (EMG and USI), and the generated torque (force plate) during movement execution, attempt, and imagination. The muscle activity detection was performed with USI, compared with EMG and torque data, and the relationship between the muscle contraction observed with USI and brain activity was explored. During the second study, biofeedback method based on USI was developed, implemented and tested for the application in muscle training. A new method of automated analysis of US videos based on comparison of pixel intensities between the frames was developed and tested against an established method using automated image segmentation and feature tracking. For the application of detecting muscle activations, both methods were in agreement, with over 99% of contraction and 88% of twitches detected consistently. The detections of muscle activity made with USI, EMG and force plate were compared showing that, in general, all three methods were able to detect muscle activations in a similar fashion. However, for weaker contractions registered during movement attempt or imagination, USI proved to detect the largest number of contractions and it also detected over 80% of all twitches during MI that were not seen with any other method. The analysis of movement related cortical potentials and event related desynchronisation showed significant differences between actual movements and imagination tasks, as well as between imagination with twitches and without any muscular activity during different phases of action. This has been linked to the role of sensory feedback, incomplete inhibition, and necessity to activate proprioceptors during MI. The presence of muscular activity resulted in different cortical processes and influenced the motor potentials associated with preparation and action execution, even if participant was not aware of activating the muscle. The development of a USI processing method capable of providing the information about muscle activity immediately after the contraction attempt led to implementing a quantitative near-real time USI biofeedback that has not been used previously. The method proved to be suitable for detecting muscle activations and to estimate the contraction intensity, which could be beneficial during motor rehabilitation. Ultrasound imaging is the most sensitive method for detection of small muscle activations and a newly developed processing techniques enables fast, reliable and automated processing. The presence of muscle contractions and twitches need to be carefully monitored, if motor imagination is to be employed in BCI paradigms and the results translated between the healthy and patient populations. USI proved to have a good potential for such application. During the training with USI based feedback, due to the sensitivity and variability of the signal, no significant improvement in controlling the muscle activation was seen between different trials and sessions. Nevertheless, some improvement occurred on a trial-to-trial basis and the individuals achieved better awareness of activating their muscle, therefore it is believed that the feedback method proposed here could benefit motor training and re-learning muscle activations after surgical procedures of different muscle structures.