Date of Award

May 2026

Document Type

Dissertation

Degree Name

Doctor of Philosophy (PhD)

Department

Biomedical Engineering

First Advisor

Enrique Alvarez Vazquez

Abstract

Parkinson’s disease is a progressive neurodegenerative disorder characterized by motor dysfunction and early cognitive decline. Diagnosis and monitoring remain largely symptom-driven, relying on clinical scales that capture disease only after significant neurodegeneration has occurred. These assessments are inherently subjective, episodic, and insensitive to early-stage pathology, limiting both timely diagnosis and the ability to track disease progression or therapeutic response. The lack of objective, quantitative biomarkers for both molecular pathology and cognitive impairment remains a major barrier to advancing precision medicine approaches in Parkinson’s disease. This dissertation approaches Parkinson’s disease from two complementary fronts: molecular detection of alpha-synuclein pathology and quantitative characterization of cognitive dysfunction.

The first approach develops a minimally invasive electrochemical impedance spectroscopy biosensor for detecting alpha-synuclein in physiologically relevant media. Surface characterization demonstrated progressive optimization of gold nanoparticle deposition, with mean surface coverage increasing from 2.20 ± 1.85% in Iteration 1 to 5.33 ± 4.13% in Iteration 3, alongside a reduction in nearest-neighbor spacing variability from 68.73% to 54.37%. Raman spectroscopy confirmed the structural stability of graphene oxide, with an ID/IG ratio of approximately 0.94 across batches and over two weeks, while alpha-synuclein-functionalized surfaces exhibited a consistent amide feature ratio of 0.9561 ± 0.0106. Electrochemical measurements demonstrated concentration-dependent impedance responses over the range of 0 to 200 pM. In phosphate-buffered saline, the strongest monotonic relationship between impedance magnitude and concentration occurred at approximately 1.06 Hz, with a Spearman correlation of 0.79, an estimated limit of detection of 50 pM, and inter-sensor variability of approximately 8%. In simulated interstitial fluid, the optimal frequency shifted to approximately 13.8 Hz, with a stronger monotonic relationship of 0.83, reduced variability of approximately 3%, and an estimated limit of detection of 200 pM. This frequency shift reflects matrix-dependent changes in interfacial dynamics. While impedance magnitude exhibited a compressed dynamic range in simulated interstitial fluid, phase analysis indicated robust sensitivity to binding-associated changes in interfacial capacitance, supporting the use of multi-feature electrochemical sensing.

The second approach develops an EEG and pupillometry feature-extraction framework integrated with a principal component analysis and support vector machine classification pipeline. A total of 108 multimodal features were extracted per participant with preservation of bin-level temporal alignment. In the working memory task, the optimal model used 14 principal components and 7 features, achieving 71% test accuracy and 63% hold-out accuracy, with an F1 score of 0.626 and balanced accuracy of 0.701, exceeding a 10,000-permutation null distribution with a p-value of 0.0013. In the sustained attention task, the optimal configuration used 14 principal components and 10 features, yielding 73% test accuracy and 70% hold-out accuracy, with an F1 score of 0.623. Several top-ranked features demonstrated independent group differences, including prefrontal alpha index with a Cohen’s d of −0.65 and FDR-adjusted p-value of 0.027, and theta–beta ratio with a Cohen’s d of 0.63 and FDR-adjusted p-value of 0.027.

Together, these approaches provide quantitative measures of alpha-synuclein burden and cognitive dysfunction. By integrating biochemical sensing with machine learning-based neural biomarker discovery, this work advances objective, scalable strategies for disease stratification and therapeutic monitoring in Parkinson’s disease.

Download

Share

COinS