Abstract

Alzheimer’s Disease (AD), the most common cause of dementia, is a dual proteinopathy exhibiting deposits of extracellular amyloid-beta (Aβ) and intracellular tau in brain tissue. Models which describe the spatial and temporal trajectories of these pathologies have been invaluable for understanding AD and have the potential to support diagnosis, disease monitoring, and treatment development. Disease staging systems, which provide protocols for assessing individual disease severity based on their extent of brain pathology, have been crucial for explaining and diagnosing AD. However, these systems have been primarily informed by post-mortem histological assessments and are not directly applicable to disease assessment in living persons. More recently, positron emission tomography (PET) imaging has enabled in vivo assessment of the distribution and magnitude of AD-related pathology. Several investigations have leveraged PET imaging to derive novel systems for AD staging. This work has addressed some of the limitations of post-mortem staging, by a) working for living persons, b) incorporating large samples, c) assessing pathology across the whole brain instead of isolated sections, and d) creating systems which capture both typical and atypical pathology patterns. However, there are several gaps and limitations in the field of data-driven neurodegenerative staging which warrant further study. In this dissertation, I describe a research program aimed at expanding and evaluating data-driven approaches for PET-based staging of neurodegenerative pathology in AD and related disorders. In Chapter 1, I apply data-driven approaches to understand the spatiotemporal progression of tau pathology based on PET data. This work identifies a low dimensionality model for describing individualized patterns of tau uptake. Furthermore, it produces and validates a staging model for assigning the severity of tau burden. In Chapter 2, I investigate tau uptake and stages of tau burden in individuals without Aβ pathology. These results demonstrate unique patterns of tau uptake in this population, supporting the notion that this population represents a separate entity from AD. In Chapter 3, I expand my data-driven staging methods by developing a model estimating stages of both Aβ and tau pathologies, providing a comprehensive framework for understanding the spatial progression of AD-specific pathology. Finally, in Chapter 4, I apply machine learning to identify subtypes of AD, each characterized by unique spatial progressions of Aβ and tau progression. The work I present integrates machine learning techniques with staging methods to support the development of empirical, reproducible systems for grading the extent of neurodegenerative pathology. I develop and demonstrate improved methods for pathological staging and reveal novel spatial and temporal trajectories of pathology in AD and related disorders.

Committee Chair

Aristeidis Sotiras

Committee Members

Arash Nazeri; Brian Gordon; Janine Bijsterbosch; Todd Braver

Degree

Doctor of Philosophy (PhD)

Author's Department

Interdisciplinary Programs

Author's School

McKelvey School of Engineering

Document Type

Dissertation

Date of Award

8-17-2026

Language

English (en)

Download

Included in

Neurosciences Commons

Share

COinS