MRI-based measures in Alzheimer's disease and related disorders : regions of interest and automated techniques

• Main Author: Barnes, Josephine
• Published: University College London (University of London) 2007
• Subjects: 616.8
• Online Access: http://ethos.bl.uk/OrderDetails.do?uin=uk.bl.ethos.487753

This thesis investigates the use of manual measures of brain structure delineation on MR scans in order to assess atrophy in dementia. It further investigates the automation of atrophy measures. A new protocol for outlining the cingulate is described and was applied to groups of Alzheimer's disease (AD) and control subjects. The application of existing hippocampal and amygdala protocols to a group of pathologically-confirmed AD, frontotemporal lobar degeneration (FTLD) and controls is detailed. This analysis shows cross-sectional measurements were useful subject-group discriminators and that patterns of atrophy within and between structures may distinguish diseases. Manual delineation of regions (cingulate and hippocampus) was extended to longitudinal studies to establish atrophy rates in groups of AD, FTLD and controls. The cingulate was shown to be at least as affected as the hippocampus by disease. Hippocampal atrophy rates from inter-scan intervals of six months were compared with measurements in the same subjects of one-year intervals. Results from studies combining cross-sectional and longitudinal hippocampal data are described. These assess the asymmetry of the structure and investigate the pre-symptomatic decrease in volume in familial AD subjects. Semi-automated techniques were performed which utilise registration of serial hippocampi to assess change longitudinally. Results show the semi-automated techniques to be reliable and consistent with manual measures. These techniques were then applied to scans from a multi-centre clinical trial and again consistency with manual measures was assessed. The generation of fully-automated template-based hippocampal segmentations is described. The approximate regions generated from the template were used to quantify the boundary shift integral and the resultant atrophy rates were compared with manual rates revealing automated measures to be consistent with manual measures. These results are put within the context of other similar studies by performing a meta analysis of hippocampal atrophy rates. Implications for diagnosis and monitoring disease progression are discussed.