Doctoral Dissertations
Date of Award
5-2023
Degree Type
Dissertation
Degree Name
Doctor of Philosophy
Major
Industrial Engineering
Major Professor
Anahita Khojandi
Committee Members
John Kobza, Jim Ostrowski, Ritesh Ramdhani
Abstract
Parkinson’s disease (PD) is a progressive, neurodegenerative disorder resulting from the loss of dopaminergic neurons. The disease is characterized by four cardinal motor symptoms, bradykinesia (slowing of movement), muscle rigidity, tremor, and postural instability/gait disorder. Early in the disease course, dopaminergic medication can effectively manage the cardinal motor symptoms for the majority of patients. The primary medication, which underpins all long-term treatment planning in PD, is levodopa (L-dopa). However, increased L-dopa typically results in increased dyskinesia (a side effect of L-dopa). As the adverse side effects of L-dopa increase in severity, patients may need advanced therapies such as deep brain stimulation (DBS), a neurosurgical treatment. Advanced therapies have been shown effective at alleviating the side effects associated with L-dopa. However, DBS requires the additional tuning of stimulation parameters to optimize treatment plans. Further, the interaction between L-dopa and DBS remains an open question. This dissertation proposes methodological and applied analysis to personalize treatment planning for PD patients throughout their disease course.
First, we optimize early-stage PD medication regimens using machine learning clustering techniques and Markov decision processes incorporating pharmacokinetic considerations for immediate release L-dopa. Further, these models are calibrated using wearable sensor data from a multi-visit cohort of PD patients. We apply reinforcement learning techniques to determine the optimal treatment plans for patient subtypes.
Second, we develop machine learning approaches to improve the efficiency of parameter tuning for DBS as well as predict potential advanced therapy candidates. We begin with a review of the literature. Then we examine a primary cohort of chronic PD patients who have received DBS. We determine the gait parameters impacted by DBS (both high and low-frequency stimulation) both with and without the influence of dopaminergic medication. This study represents the largest study of its type, focusing on low-frequency stimulation in combination with medication.
Finally, we examine the role of machine learning and dimensionality reduction in the prediction of gene expression. Genetics influence a variety of chronic diseases, including Parkinson’s disease. This work sought to improve the computational efficiency of gene expression prediction models for use in the progression modeling of several chronic diseases.
Recommended Citation
Watts, Jeremy, "Machine Learning and Decision Making to Optimize Treatment Planning in Parkinson’s Disease. " PhD diss., University of Tennessee, 2023.
https://trace.tennessee.edu/utk_graddiss/8106
Included in
Artificial Intelligence and Robotics Commons, Biostatistics Commons, Data Science Commons, Diagnosis Commons, Industrial Engineering Commons, Medical Genetics Commons, Neurology Commons, Neurosciences Commons, Neurosurgery Commons, Other Pharmacy and Pharmaceutical Sciences Commons