Computational Modeling of Spinal Cord Stimulation for Inspiratory Muscle Activation and Chronic Pain Management

Abstract

Spinal cord stimulation (SCS) is a neuromodulation technique that applies electrical stimulation to the spinal cord to alter neural activity or processing. While SCS has historically been used as a last-resort therapy for chronic pain management, novel applications and technologies have recently been developed that either increase the efficacy of treatment for chronic pain or drive neural activity to produce muscular activity/movement following a paralyzing spinal cord injury (SCI). Despite these recent innovations, there remain fundamental questions concerning the neural recruitment underlying these efficacious results. This work evaluated the neural activity and mechanisms for three SCS applications: both conventional SCS and closed-loop SCS for pain management, as well as ventral, high frequency spinal cord stimulation (HF-SCS) for inspiratory muscle activation following a SCI. I developed computational models to both predict the neural response to SCS and explore factors influencing neural activation. Models consisted of three components: a finite element model (FEM) of the spinal cord to predict the potential fields generated by stimulation, biophysical neuron models, and algorithms to apply time-dependent extracellular potentials to the neuron models and simulate their response. While this cutting-edge modeling methodology could be used to predict neural activity following stimulation, it was unclear how anatomical and technical factors affected neural predictions. To evaluate these factors, I designed an FEM of a T9 thoracic spine with an implanted electrode array. Then, I sequentially removed details from the model and quantified the changes in neural predictions. I identified several factors with large (>30%) effects on neural thresholds, including overall electrode impedance (for voltage-controlled stimulation), the electrode position relative to the spine, and dura mater conductivity. This work will be invaluable for basic science and clinical applications of SCS. Next, I developed a canine model to evaluate T2 ventral HF-SCS for inspiratory muscle activation after an SCI. This model infrastructure included two neuron populations hypothesized to lead to inspiratory behavior: ventrolateral funiculus fibers (VLF) leading to diaphragm activation and inspiratory intercostal motoneurons. With this model, I predicted robust VLF and T2-T5 motoneuron recruitment within the experimental range of stimulation. I used this model framework to optimize several design parameters related to HF-SCS for inspiration. The optimal lead design parameters were evaluated via in vivo experiments, which found excellent agreement with model predictions. This work expands our mechanistic understanding of this novel therapy, improves its implementation, and aids in future translational efforts towards human subjects. Finally, I developed a computational model to evaluate closed-loop SCS for chronic pain management. This work characterized the neural origins of the evoked compound action potential (ECAP), the controlling biomarker of closed-loop stimulation. This modeling work showed that ECAP properties depend on activation of a narrow range of axon diameters and quantified how anatomical and stimulation factors (e.g., CSF thickness, stimulation configuration, lead position, pulse width) influence ECAP morphology, timing, and neural recruitment. These results improve our mechanistic understanding of closed-loop stimulation and neural recruitment during SCS. In summary, this dissertation work improves the methodology, validation, and applications of computational models of SCS. It also has direct applications to the clinical/pre-clinical implementation of SCS and may be invaluable for expanding the utility and efficacy of several treatments. The improved mechanistic understandings of neural activation described in this work may also aid in the future development of these therapies.