Quantifying structural changes in the ageing brain from magnetic resonance imaging

Understanding the ageing process is of increasing importance to an ageing society and one aspect of this is investigating what role the brain has in this process. Cognitive ability declines as we age and it is one of the most distressing aspects of getting older. Brain tissue deterioration is a significant contributor to lower cognitive ability in late life but the underlying biological mechanisms in the brain are not yet fully understood. One reason for this is the difficulty in obtaining accurate measures of potential ageing-related brain biomarkers. The chapters in this thesis explore the difficulties of quantifying brain changes in the ageing brain from Magnetic Resonance Imaging (MRI), and how the changes identified are related to cognition in later life. The data was acquired as part of the second wave of the longitudinal Lothian Birth Cohort 1936 study in which 866 people aged 73 years, returned for cognitive and medical assessment. At this stage of the study 702 underwent MR imaging resulting in 627 complete datasets across all testing. The entire data, a randomly chosen subset of 150 and 416 freely available data were used to investigate global and regional measurement methods in older brains and how the resultant measurements related to cognitive performance. Furthermore the presence of early life cognitive data in the form of a general intelligence test sat at age 11, served as an indicator of cognitive ability prior to the potential influence of the ageing process. The chapters concerning global measures at first establish, that a measure of intracranial volume (ICV) serves as both a way of correcting for individual differences in brain size between participants and as a proxy premorbid measure of brain size. The analysis, utilising freely available cross-sectional MRI data (http://www.oasis-brains.org) revealed that ICV differed very little between 18-28 year olds and 84-96 year olds where as total brain tissue volume (TBV) differed by 14.1% between the two groups, which was more than twice the standard deviation across the entire age range (18-96 years). Second a validated, reliable method for measuring ICV was investigated using 150 people randomly chosen from the LBC1936 study. Automated and semi-automated methods were validated against reference measurements the results of which showed that common ageing features make automated and semi-automated methods that do not have an additional manual editing step, ineffective at producing accurate ICV measurements. This analysis also highlighted the need to employ additional spatial overlap assessment to volumetric comparison of measurement methods to reduce the effect of false-positives and false-negatives skewing apparent discrepancies between methods. Using the information gained here ICV and TBV from the entire LBC1936 cohort were analysed in a structural equation model, alongside cognitive ability measures at both age 11 and age 73. We found that TBV was a stronger predictor of later life cognitive ability, after accounting for early life ability, but that a modest association remained between ICV and late life cognition. This suggests that early life factors pay a role in how well we age, though the relationship is complex. The regional measures chapters look at two brain regions commonly associated with ageing, the hippocampus and the frontal lobes. Measuring either of these brain regions in large samples of healthy older adults is challenging for many reasons. The hippocampus is small and as with all brain regions shows greater variation in older age, this makes employing automated methods that have the advantage of being fast and reproducible difficult. Following the results of our systematic review of automated methods for measuring the hippocampus, the two most commonly used and available automated methods were validated against reference standard measurements. The results indicated that although automated methods present an attractive alternative to laborious manual measurements they still require manual editing to produce accurate measurements in older adults. The modified strategy employed across the LBC1936 was to use an automated method and then manually edit the output; these segmentations were used to investigate the potential of multimodal image analysis in clarifying associations between the hippocampus and cognitive ability in old age. The analysis focused on associations between longitudinal relaxation time (T1), magnetization transfer ratio (MTR), fractional anisotropy (FA) and mean diffusivity (MD) in the hippocampus and general factors of fluid intelligence, cognitive processing speed and memory. The findings show that multi-modal MRI assessments were more sensitive than volumetric measurements at detecting associations with cognitive measures. The difficulty with producing a relevant frontal lobe measure was made apparent when the result of a large systematic review looking at the manual protocols used revealed 19 methods and 15 different landmarks had been employed. This resulted in an analysis that took the 5 most common boundaries reported and applied them to 10 randomly selected participants from the LBC1936. The results showed significant differences between the resultant volumes, with the smallest measurement when using the genu as the posterior marker representing only 35% of the measurement acquired using the central sulcus. The results from the studies presented in this thesis strongly highlight the need to develop age specific methods when using brain MRI to study ageing. Furthermore the implications of using unstandardised protocols, making assumptions about a methods performance based on validation in younger samples and the need to account for early life factors in this area of research have been made clearer. Studies building on these findings will be beneficial in elucidating the role of the brain in ageing