Use this URL to cite or link to this record in EThOS:
Title: Decoding the neural basis of age-related decline : combining pattern recognition and diffusion magnetic resonance imaging
Author: Parker, Richard Garry
ISNI:       0000 0004 7963 9148
Awarding Body: King's College London
Current Institution: King's College London (University of London)
Date of Award: 2018
Availability of Full Text:
Access from EThOS:
Access from Institution:
Pattern recognition techniques are gaining popularity as a tool for probing biomedical image data - affording i), accurate predictions of clinically-relevant variables (presence of disease, degree of cognitive impairment, etc) at the single-subject level and ii), novel means of visualising brain structure-function mappings. This thesis constitutes a rigorous examination of the utility of pattern recognition for understanding brain structural changes in later-life. We focus upon the application of pattern recognition techniques to diffusion MRI (D-MRI) - a technique which elucidates tissue microstructural properties not measurable with other non-invasive scanning techniques - but which to date has received only modest attention as a data source for pattern recognition analysis (compared to conventional T1-MR sequences). In the first experimental chapter, we replicate several previous experiments by showing that pattern recognition methods can be applied to whole-brain D-MRI data in order to diagnose the mild cognitive impairment (MCI) condition, as well to predict subject chronological age. This experiment builds upon previous efforts by demonstrating that choice of image preprocessing strategy can markedly influence the ability to predict variables of interest from diffusion scans, and we also provide novel evidence that pattern recognition can be used to decode memory functioning from D-MRI scans of healthy older adults. In the second experimental chapter we examine prediction of experimental variables from specific brain white matter regions (rather than whole-brain representations), as targeted pattern recognition analysis can provide more accurate/interpretable results. From the extant literature, we hypothesised that diffusion measurements extracted from the major fibre pathways of the limbic system - the uncinate fasciculus, fornix and parahippocampal cingulum - would carry enough information to allow for accurate predictions of cognitive impairment in samples of older adults. Drawing on elements from several recent publications, we present a pipeline for conducting pattern recognition analysis upon detailed, 'along-tract' representations of these structures, attained using the diffusion tractography technique. The method we propose confers several theoretical advantages over traditional tract-specific D-MRI analysis approaches. In the final experiment 5 we investigate the use of pattern recognition-based age prediction models (trained upon the D-MRI scans of healthy adults) as a means of characterising brain structural alterations in MCI - the so-called BrainAGE technique (Franke, et al., 2010). After constructing normative models from a large, multi-scanner cohort, we identify (for the first time) that diffusion-based age predictions are systematically over-estimated in individuals with MCI by around 2.8-4.4 years, complementing previous BrainAGE studies using T1-MRI, and suggesting that an 'accelerated-ageing'-type mechanism at least partially accounts for the white matter alterations that occur in MCI. We believe that this thesis contributes significantly to the study of cognitive ageing, and we hope that our results can help guide the design of future diffusion MRI-based pattern recognition experiments.
Supervisor: O'Sullivan, Michael ; Marquand, Andre Frank Sponsor: Not available
Qualification Name: Thesis (Ph.D.) Qualification Level: Doctoral
EThOS ID:  DOI: Not available