| Summary: | Many drugs are under development to treat Alzheimer's disease (AD), the prevalent cause of dementia. Cognitive tests are the established way of assessing drug efficacy. More sensitive, quantitative measures, biomarkers, are desirable for faster assessment of efficacy from smaller cohorts. AD is characterized by amyloid plaques, neurofibrillary tangles, neuronal cell death, memory deficits and brain volume loss detectable using longitudinal neuroMRI. This thesis focuses on acquisition artefacts which hinder the use of longitudinal structural MRI as an accurate, precise biomarker. It investigates change in artefact magnitude with field strength, prevalence of artefacts in clinical trials, influence of artefacts on quantification of atrophy, and correction of carotid flow artefacts. A prospective study was used to quantitatively compare pulsatile flow artefact on images from a 3T scanner with those from a standard clinical 1.5T scanner. The artefact was significantly worse at the higher field strength (p < 0.01). In images from 837 clinical trial subjects, contrast issues, motion, and temporal lobe (occurring in up to 41% of scans) were the prevalent artefacts. The temporal lobe, particularly important in AD research, was affected predominantly by pulsatile flow artefact (60-70% of temporal lobe artefacts). Moving trials to higher field strengths may significantly increase these artefacts' occurrence. I investigated the effects of artefact on computerized brain atrophy measurement techniques: the Boundary Shift Integral (BSI) and SIENA. I developed methods of simulating various types and severities of motion and pulsatile flow artefacts. Images simulated with known atrophy and artefacts were used to test the techniques. Rotational motion > 3 at follow-up can result in apparent atrophy, of the order of AD relative to normal annual ageing, (1.5%), using SIENA. Increasing motion decreases measured atrophy using BSI: > 3 can obscure atrophy of 1.5%. An algorithm to remove pulsatile artefact was devised, implemented and tested on scans with simulated artefacts.
|
|---|
