Learning to move: exploring the neural dynamics of motor learning

This thesis reflects an interdisciplinary research project aimed, in part, to address the question: “How can Brain-Computer Interface (BCI)-based motor rehabilitation be optimized?” It also sought to expand the scientific understanding of the neural processes associated with motor learning. Specifically, we investigated three neural correlates of motor control – Event-Related Desynchronization (ERD), Motor Related Cortical Potential (MRCP), and the temporal evolution of Corticospinal Excitability (CSE) – with the goal of characterizing their interactive relationships and identifying any dynamic changes associated with motor learning. Despite the increasing use of these neural markers in contexts such as BCI-based post-stroke motor rehabilitation, the functional connectivity, and dynamics of these markers during the process of motor learning remain poorly understood. Especially in relation to improving performance in motor skill-based tasks. This research aims to clarify these dynamics to optimize BCI setups and advance motor rehabilitation. Though this thesis is framed around improving motor rehabilitation, its exploratory nature necessitated working exclusively with healthy participants, serving as a foundational step toward future clinical applications. The first study covers our analysis of data by Daly et al. (2018), to further explore their suggested time dependent relationship between ERD and CSE. The second and third studies focus on two renditions of a new motor learning experiment, one to behaviorally validate the design and the other to collect EEG and TMS response data to replicate and expand on our findings in the data by Daly et al. (2018). Taken together, our results indicate the temporal evolution of CSE, as measured by MEP amplitude in the 2 seconds leading up to Movement Onset, follows an S-like wave (third-degree polynomial). Where an initial increase in amplitude is followed by a decline after which it once again changes direction to strongly move upward. While ERD measures showed potential for predicting this changing CSE timeline, the cubic relationship between CSE and time did not extend to describe the relationship between CSE and ERD. Instead, preliminary insights suggest that the CSE-ERD relationship is unstable, implying that their connection is more likely correlational—based on a shared temporal progression relative to MOn—rather than functionally dependent. Furthermore, both CSE and ERD significantly changed with motor learning, with ERD power decreasing further and CSE amplitude reducing following early-stage learning but further unchanged. Despite our inability to quantify a stable relationship between ERD and CSE, our findings clearly demonstrate that learning affects the reliability of deriving CSE from ERD, as their relationship remains unstable over time. These findings highlight that ERD dynamics vary between individuals and frequency bands, discouraging the use of fixed percentage thresholds to derive optimal excitability. Additionally, because learning alters ERD power, it must be accounted for when using ERD to infer CSE dynamics. Future research should focus on refining methodologies to better understand these dynamic interactions and their implications for motor control and BCI applications. This newfound understanding of how motor learning occurs in the brain is of interest to further our understanding of how BCI-based rehabilitation works and will help to optimize the development of BCI-driven motor recovery paradigms.