Shawa, Zeena; (2025) Modelling Heterogeneity in Neurodegenerative Disease with MRI. Doctoral thesis (Ph.D), UCL (University College London).

Preview

Text ModellingHeterogeneityinNeurodegenerativeDiseasewithMRI_ZeenaShawa.pdf - Accepted Version Download (54MB) | Preview

Abstract

Neurodegenerative diseases such as Alzheimer’s disease (AD), Parkinson’s disease (PD), and Lewy Body Dementias (LBDs) pose a growing global healthcare challenge, particularly as populations age. Despite their prevalence, understanding of their progression and treatment remains limited, largely due to their clinical and biological heterogeneity. However, the increasing availability of large multimodal datasets presents new opportunities to apply statistical modelling and machine learning techniques to uncover meaningful disease progression signatures and biomarkers. This thesis explores how magnetic resonance imaging (MRI), a widely used, non-invasive modality, can be leveraged alongside these approaches to investigate neurodegenerative disease heterogeneity across three contexts: subtyping, hippocampal subfields, and ethnicity. First, using the largest T1-weighted MRI dataset in PD, I applied a state-of-the-art disease progression model to identify four spatiotemporal atrophy subtypes. Longitudinal clinical data revealed that differences in progression were primarily between those with no detectable atrophy and those with atrophy, suggesting that severity, rather than atrophy location, distinguishes subtypes. Next, I studied hippocampal subfields in PD and LBD. I found that specific subfield atrophy was associated with cognitive decline, with significant differences between diagnostic groups. While our findings align with some previous studies, the overall variability across literature highlights the underlying heterogeneity of these conditions. Finally, I assessed the fairness and generalisability of a brain age model, originally trained on an ethnically uniform population and used as a biomarker for dementia risk, on an ethnically diverse dataset. While the model performed comparably across aggregated White and Non-White groups, performance declined for individual minority groups, with minimal intersectionality effects observed. These findings underscore the need for more representative training data and rigorous evaluation in under-represented populations. Together, these studies advance our understanding of neurodegenerative disease progression and could inform clinical trial recruitment and treatment development, ultimately improving outcomes for diverse patient populations.

Download activity - last month

Export as CSVXMLJSON

Download activity - last 12 months

Export as CSVJSONXML

Downloads by country - last 12 months

Export as XMLJSONCSV

Archive Staff Only

View Item