Early gait rehabilitation is a critical priority for individuals with cerebral palsy (CP), as neurologic injury in development alters motor control and can lead to a series of progressive impairments which may limit independent mobility. However, given the inherent heterogeneity of brain injury, developing strategies that can consistently improve function in CP has proven challenging;​ to date, outcomes from traditional interventions elicit measurable gait changes in approximately 50% of cases and often fail to directly target motor control. Biofeedback training, whereby individuals are provided with real-time information on gait performance, is a promising alternative to traditional interventions, as systems can be flexibly tuned to provide individualized and task-specific practice. Yet, early studies using biofeedback in CP have reported many of the same limitations as traditional interventions, outcomes are often unsatisfactory and motor control is largely unchanged. As such, in order for biofeedback systems to be reliably integrated into clinical care, there is a need to understand and improve upon this heterogeneity. The goal of this dissertation is, therefore, two-fold:​ (1) to understand the factors that modulate motor control during gait which will help inform the extent to which it can be effectively targeted with biofeedback, and (2) to characterize how the design and implementation of biofeedback systems may influence outcomes.

In order to successfully alter motor control in individuals with CP using biofeedback, there is a need to first understand which factors may underlie such measures. Motor control following neurologic injury is simplified relative to nondisabled individuals, defined by greater co-activity of antagonist muscle groups during walking. While these simplified strategies are hypothesized to be the direct result of decreased involvement of supraspinal pathways in the production of movement, they may also partially reflect the biomechanical constraints imposed by pathologic gait patterns. To isolate the effects that altered gait has on motor control, we evaluated whether motor control strategies changed when nondisabled adults emulated common gait patterns in CP (e.g., crouch gait, equinus). During emulation, motor control did not change significantly from baseline and did not align with motor control strategies observed in a cohort of individuals with CP walking in the same patterns. Together, this bolsters evidence that simplified motor control in CP is a direct effect of the primary neurologic injury and may, therefore, not be significantly improved by solely targeting gait kinematics via intervention.

The observed stability of motor control during gait pattern emulation aligns with prior studies which have demonstrated that motor control strategies are robust across changing biomechanical contexts such as cycling, running, and walking with partial body-weight support. Although collectively this suggests that motor control may be inflexible and, therefore, not modifiable with biofeedback training, this work has been largely constrained to a subset of achievable walking configurations, limiting the overarching conclusions which can be drawn. As such, we evaluated the extent to which nondisabled adults could dynamically modify motor control using a custom-built motor control-based biofeedback system to guide broad gait pattern exploration. The resultant data set enabled us to model how motor control changed as a function of imposed biomechanical constraints. We found that individuals could consistently simplify motor control in response to biofeedback by altering their gait mechanics, predominantly at the knee and ankle. This work suggests that motor control may be modified during biofeedback training, but that the extent of that modification is contingent on the parameter targeted.

While the results of these studies ultimately improve understanding of the extent to which motor control may be modified during walking, other factors may influence response to biofeedback training that warrant additional investigation. In particular, although a variety of biofeedback systems have been developed for use in CP, it is unclear how the choice of modality used to communicate error may affect response. We compared how individuals with CP adapted motor control in response to audiovisual and sensorimotor biofeedback paradigms when administered independently and in combination;​ both systems were designed to directly target plantarflexor recruitment. We found that individuals were able to rapidly modify plantarflexor activity using each of the biofeedback paradigms, but that they primarily relied on the audiovisual system to guide movement correction. This indicates that individuals with CP may differentially prioritize distinct forms of biofeedback which must be considered in the future design of systems.

Individuals’ capacity to retain improvements with repeated biofeedback training and transfer in-session gains beyond the training paradigm may also underlie the heterogeneity of outcomes. Extending experimental frameworks in motor adaptation, we evaluated how motor control changed as a function of multi-session practice with audiovisual and sensorimotor biofeedback and whether changes were retained once biofeedback was removed. We observed that the extent to which individuals retained adapted gait patterns was contingent on the magnitude of in-session gains. This suggests that an individual’s adaptation capacity may dictate their response to biofeedback and may, therefore, be valuable to consider when selecting optimal candidates for training.

Biofeedback training is an exciting avenue to personalize rehabilitation for individuals following neurologic injury and may serve as a non-invasive alternative to traditional interventions. By developing analytical frameworks to better understand motor control and evaluate how individual capacity and system design choices influence response to biofeedback, this dissertation lays the foundation to advance the design of clinically translatable systems which can support long-term, independent mobility in CP.