Operant conditioning of monosynaptic spinal reflexes and its computational modeling and simulation

Spasticity, or more specifically hyperreflexia, is a common impairment following neurological injury such as stroke. Current clinical interventions aimed at reducing rectus femoris (RF) hyperreflexia have shown modest effect but entails side effects and limited clinical evidence. My previous research has shown that RF hyperreflexia is associated with reduced knee flexion in people post-stroke with Stiff-Knee gait (SKG). I posit that reducing RF hyperreflexia should improve walking following SKG after stroke. I developed a non-pharmacological procedure using operant H-reflex conditioning of the RF, which allows the patient to self-modulate one’s spinal reflex activity, elicited via electrical stimulation on peripheral nerve. With current evidence that operant H-reflex conditioning enhances gait function in individuals with SCI, I conducted a proof-of-concept study to examine the feasibility of this procedure on the RF for healthy and post-stroke individuals. Operant conditioning of neural activation has a high incidence of non-responders, and delineating the explicit response to feedback can help determine why some individuals may not respond to neurofeedback training. I developed a simulated operant H-reflex conditioning neurofeedback environment that separated the ability to self-regulate the neurofeedback signal from its perception by using an explicit, unskilled visuomotor task. Main outcomes indicated that biological variability modulates performance and operant strategy depending on the feedback type. While previous results provided a holistic view of the effect of feedback parameters on overall performance and operant strategy, the next approach focused on determining whether such decisions could be predicted based on feedback on a trial-by-trial basis. I observed that the feedback sensitivity was modulated by biological variability and reward threshold. I used computational models to investigate the best estimate of learning resulting in feedback-weighted averages of previous decisions. This thesis introduces a novel simulated operant H-reflex conditioning environment that serves as a simple and robust model to quickly examine learning mechanisms, optimize learning, and potentially identify non-responders.