Analysis of functional neuroimaging : effects of preprocessing on outcome measures with particular reference to single subject response

Functional neuroimaging analysis is generally applied to studies of groups of subjects. The most commonly used imaging modality is functional magnetic resonance (fMRI) and most software packages are designed mainly for the analysis of fMRI data of task-based studies. The same software and approaches to analysis are also applied to other modalities, such as positron emission tomography (PET) and single photon emission computed tomography (SPECT), and to different study types, such as the detection of abnormalities in individual brain scans by comparison with a normal database. The validation and optimization of individual parts of processing and complete analyses usually relies on the use of well-characterised human, modified human or simulated data. Work presented in this thesis employs well-characterized human data for the optimization of a study involving the administration of high doses of a drug to subjects. Imaging data was acquired using two complementary pharmacological magnetic resonance (MR) techniques - arterial spin labeling (ASL) and blood oxygen level dependent (BOLD). After optimization, the concordance between the two techniques and overall sensitivity was shown to be improved. The application of optimization techniques with respect to areas such as ageing, image registration and count scaling has been demonstrated to markedly improve the reliability and sensitivity of the voxel-based techniques. Because there are issues surrounding the comparability of datasets acquired on different cameras, using a printed subresolution phantom has been shown to be useful in the investigation of the imaging characteristics of different systems. In conclusion, this work has demonstrated the importance of optimization and simulation across a number of different studies with a particular focus on single subjects. The range, scope and complexity of neuroimaging studies can be expected to increase in the future in response to the ageing population, increasing computer power and improved understanding for disorders affecting the brain.EThOS - Electronic Theses Online ServiceGBUnited Kingdo