Dementia is a syndrome that causes severe disorders of higher brain functions and thus greatly affects patient’s everyday activity. It is often caused by Alzheimer’s disease (AD), a common neurodegenerative brain disease. Due to similar clinical signs the diagnosis of neurodegenerative brain diseases is often challenging. Therefore an objective biomarker is needed to confirm the presence of the disease in its early stages.
To improve the diagnosis of neurodegenerative brain diseases positron emission tomography (PET), a functional brain imaging technique, is widely used. Prior to scanning radioactive tracer fluorodeoxyglucose (FDG), a glucose analogue, is intravenously applied to the subject. FDG is then distributed within subject’s body and brain, so that the tracer density is high in the brain areas with high metabolic activity and low in the areas of low metabolic activity. Qualitative analysis of FDG-PET brain images can be used to discover changes in brain metabolic activity caused by AD before they are visible on structural brain images. What is more, by utilising Scaled Subprofile Model/Principal Component Analysis (SSM/PCA) it is possible to derive a characteristic brain metabolic pattern of Alzheimer’s disease (ADRP) from the FDG-PET brain images of AD patients and healthy control participants (CN).
There are several parameters affecting the shape and the clinical applicability of ADRP. This thesis explores the effect of both technical quality and number of FDG-PET brain images used in the pattern derivation process on the shape, diagnostic capability and clinical applicability of ADRP. Therefore an overview of AD and PET is presented, followed by a description of image selection and pre-processing. Moreover, the procedures of pattern derivation and validation are described. Then three different ADRPs are derived and validated by employing various qualitative and quantitative statistical methods. Finally, three hypotheses defined early in this thesis are tested by analysing results obtained during pattern validation.
|