Exploring Alzheimer's Disease Stages: Insights Through t-SNE and UMAP Visualizations

Date of Award

12-2024

Degree Name

Master of Science

Department

Electrical and Computer Engineering

First Advisor

Ikhlas Abdel-Qader, Ph.D.

Second Advisor

Saad Shebrain, M.D.

Third Advisor

Simin Masihi, Ph.D.

Keywords

Alzheimer's disease, biomarkers, cognitive decline, neuroimaging, t-SNE (t-Distributed Stochastic Neighbor Embedding), UMAP (Uniform Manifold Approximation and Projection)

Access Setting

Masters Thesis-Abstract Only

Restricted to Campus until

12-1-2026

Abstract

This thesis applies advanced visualization techniques to analyze patterns within the Alzheimer’s Disease (AD) Neuroimaging Initiative (ADNI) dataset, involving multimodal neuroimaging data, clinical assessments, and genetic information. Dimensionality reduction methods, such as t-distributed Stochastic Neighbor Embedding (t-SNE), Uniform Manifold Approximation and Projection (UMAP), were utilized to manage the data. This research focuses is on the visualization and interpretation of complex data structures across various cognitive states: AD, Early Mild Cognitive Impairment (EMCI), Late Mild Cognitive Impairment (LMCI), along with Cognitively Normal (CN). The results show that t-SNE is effective in capturing fine-grained local clusters and distinguishing cognitive subgroups.

In contrast, UMAP offers a comprehensive view of the local and global data structures, enabling the visualization of transitions between cognitive states. They reveal relationships within the data, providing insights into disease progression and potential biomarkers for diagnosing AD, thereby aiding in early detection. This work highlights the utility of visualization in uncovering patterns within high-dimensional medical data, enhancing analysis, assessing the effectiveness of the t-SNE algorithm in identifying clusters and patterns within the ADNI data in two and three dimensions. The results highlight the weaknesses and strengths of both algorithms in terms of accuracy, interpretability, and computational efficiency.

This document is currently not available here.

Share

COinS