Schroder, Anna; (2024) A Framework for Pathology-Informed Optimisation of the Tractography-Estimated Structural Connectome. Doctoral thesis (Ph.D), UCL (University College London).

Preview

Text AnnaSchroder_Thesis_with_corrections2.pdf - Accepted Version Download (105MB) | Preview

Abstract

Over the past 25 years tractography has become a key tool in neuroscience, enabling virtual reconstruction of white matter bundles and construction of the brain’s anatomical connectivity, or connectome, in vivo and non-invasively. Despite the promise and insights that tractography has provided, significant challenges and questions arise over its utility owing to high rates of connectivity errors. In particular, an overwhelming number of false positive connections are consistently identified in tractography-estimated connectomes. Recent Alzheimer’s disease modelling approaches have focussed on utilising the brain’s connectivity networks to predict regional accumulation of pathology. Comparison of these predictions to measured data enables quantitative evaluation of hypothetical disease mechanisms, ultimately aiming to inform viable strategies for disease intervention. Connectivity-based models assume mechanisms of disease either based on characteristics of the connectome itself, or they mediate pathology spread through the connectome. Therefore, they are highly dependent on the accuracy of the brain’s connectivity estimate, and high connectome error rates will likely disrupt these models. This thesis evaluates tractography modelling errors and considers how models of disease pathology can play a role in correcting tractography-based estimates of brain connectivity. I present a computational framework to enable this and evaluate improvements to both connectivity estimates and predictions of regional pathology. The first contribution of this thesis focusses on tractography modelling errors. I provide a framework to identify false positive connections in the structural connectome by considering consistency of tractography streamlines with simulated diffusion images, where diffusion images are simulated via forward modelling. Specifically, I analyse the potential that each connection is a false positive, referred to as the false positive potential (FPP). I show how the FPP can be incorporated into a model of disease pathology and demonstrate the observed improvements to pathology prediction when accounting for the FPP. The second contribution considers the role pathology could play in optimising the structural connectome. Here, I exploit links between brain connectivity and propagation of neurodegeneration to constrain the tractography-estimated structural connectome. This approach yields improvements to both connectome accuracy and pathology prediction. Furthermore, I show the best results can be found when optimisation is informed by both pathology and the FPP, highlighting the need for the FPP. This thesis explores the novel idea of incorporating models of disease pathology into our estimation of the structural connectome. This work has implications in both estimates of connectivity, and in predictions of pathology accumulation. While we focus on Alzheimer’s disease here, this work has application to a range of neurodegenerative diseases.

Download activity - last month

Export as JSONCSVXML

Download activity - last 12 months

Export as JSONCSVXML

Downloads by country - last 12 months

Export as JSONXMLCSV

Archive Staff Only

View Item