| Deep brain stimulation (DBS) is an effective therapy for ameliorating motor symptoms associated with Parkinson's disease (PD) and essential tremor (ET). DBS is a surgical procedure in which implanted electrodes are placed near brain structures believed to cause pathological motor disorder behaviors. Subcortical brain structures in thalamus and basal ganglia are typical DBS targets for PD and ET therapy, and modern DBS systems allow customization of therapeutic stimulation patterns through control of stimulation location, amplitude, pulsewidth, and frequency in order to optimize therapy on a patient-specific basis. Despite the technological capabilities of implanted DBS systems, current clinical implementation is limited by a scarcity of DBS specialty clinics and variable, subjective clinician performance in assessing patient symptoms and systematically selecting optimal DBS settings for therapy, a process known as programming. Compounding this problem is a dated and inefficient healthcare system that requires patients to travel to clinic to receive DBS programming updates, oftentimes at great expense both to patients and healthcare providers.;In this dissertation, we investigate and demonstrate experimental data-driven methodologies and algorithms aimed at improving DBS therapy for PD and ET patients. Particularly, we develop adaptive DBS systems by coupling low-cost commercial sensors with stimulation control and demonstrate the therapy potential through clinical experiments with human subjects diagnosed with PD and ET. We make use of a commercially available wearable smartwatch and show that clinical ratings for tremor and bradykinesia can accurately be classified on the basis of inertial measurement unit (IMU) data. We embed this feature in a software algorithm for automated DBS programming that is demonstrated to perform at the level of expert clinicians in selecting optimal DBS settings for therapy in PD and ET patients. We further use this platform to investigate neural correlates and biomarkers of effective DBS in ET patients using recorded local field potentials (LFPs) from an electrocorticographic (ECoG) strip placed over motor cortex. Scalable methods and algorithms for automatically programming DBS when considering a large number of candidate DBS settings are also developed in detail. Finally, we address the problem of closed-loop DBS, in which stimulation is delivered in a responsive manner to patient symptoms, by formulating a model-based approach using system identification experiments, hybrid systems modeling, and predictive control.;The results presented in this dissertation may improve patient care, patient quality of life, and the cost-effectiveness of DBS therapy for PD and ET patients. Additionally, the novel results presented on neural stimulation optimization may be useful more broadly in the growing medical and research fields of neuromodulation and bioelectronic medicine. We conclude with a discussion concerning how to integrate this data-driven approach to DBS therapy in modern healthcare systems, and we discuss the relevance of these results in light of anticipated technological developments in DBS systems. |
