Electronic Theses and Dissertations

Date

2025

Document Type

Dissertation

Degree Name

Doctor of Philosophy

Department

Biomedical Engineering

Committee Chair

Amy Curry

Committee Chair

Vida Abedi

Committee Member

John Williams

Committee Member

Richard Smith

Abstract

Healthcare is currently in a transitional stage between demonstrating the potential of clustering algorithms in personalized medicine and their widespread adoption. With recent advancements in unsupervised data analysis, clustering methods have become increasingly important in disease subtyping. However, the design, development, optimization, and implementation of automated subtyping systems require further maturity before they can be fully integrated into healthcare systems. This dissertation investigates the application of clustering algorithms in creating automated subtyping systems and highlights the current challenges and limitations that must be addressed to facilitate clinical translation into personalized healthcare. Moreover, K-mean and density-based clustering were used to identify different subtypes of glaucoma disease. Specifically, we identified and characterized a subtype of ocular hypertensive eyes in the ocular hypertension treatment study (OHTS) that better responded to topical hypertensive medications. Even though the eyes in this cluster had a worse baseline mean deviation (MD) compared to eyes in all other clusters (-2.2±0.1 dB versus +0.47±1.2 dB, P < 0.001) and patients were older (64.3 ±9.5 versus 54.7 ±8.94 years old, P < 0.001), they had the best response to topical medication (MD rate of +0.04 dB/year versus -0.09 dB/year, P=0.03). Additionally, active eyes in this cluster had a significantly better MD rate than all other active eyes (MD rate of +0.04 versus -0.07 dB/year, P < 0.001). The essential characteristics of eyes in this cluster were a less frequent family history of glaucoma (P=0.006), higher frequency of heart disease (P< 0.001), and higher blood pressure (P < 0.001). We also identified a subtype of glaucomatous patients with a higher risk of rapid visual field (VF) progression based on information from the onset visit. Bayes identified that MD threshold of -3.9 dB can separate eyes in this cluster from eyes in the other clusters at onset visit. This may suggest that eyes with MD worse than -3.9 dB at the onset visit are at higher risk of rapid progression and subsequent vision loss. The identified MD threshold may assist clinicians to stratify patients based on their risk of future vision loss.

Comments

Data is provided by the student.

Library Comment

Dissertation or thesis originally submitted to ProQuest.

Notes

Open access

Download

Share

COinS