Beta 1

Title Multimodal analysis of the serotonergic neurotransmitter system
Author Rasmussen, Peter Mondrup
Supervisor Hansen, Lars Kai (Intelligent Signal Processing, Department of Informatics and Mathematical Modeling, Technical University of Denmark, DTU, DK-2800 Kgs. Lyngby, Denmark)
Institution Technical University of Denmark, DTU, DK-2800 Kgs. Lyngby, Denmark
Thesis level Master's thesis
Year 2008
Abstract Aim: In this thesis different aspects regarding multimodal analysis at voxellevel within the field of neuroimaging was investigated. The focus was on preprocessing of image data, data modeling, data analysis, and interpretation of results. Materials/methods: A review of the theory behind both the univariate general linear model (GLM) and the multivariate approaches, principal component anlaysis (PCA), partial least squares (PLS) and canonical correlation analysis (CCA) was given. Furthermore, the multivariate methods were formulated in a computationally efficient way that facilitates their application to large neuroimaging data sets. A novel thresholding strategy for quantified PET images was proposed as an additional part of the preprocessing pipeline. This may reduce the impact of quantification noise to a voxel-level analysis of quantified PET images. An extensive analysis of a data set acquired from healthy subjects was performed. The data set originated from the CIMBI project, and included positron emission tomography (PET) images of the serotonin receptor subtype 5HT2A and the serotonin transporter (5HTT), as well as structural magnetic resonance (MR) images. Results: In a conventional unimodal analysis no associations between the five NEO-PI-R personality dimensions and 5HT2A receptor binding (BP) were observed. A strong global negative correlation between 5HT2A BP and age was observed. The widespread spatial distribution in this negative relationship was dramatically reduced when gray matter (GM) concentrationwas included as a voxel-wise covariate. In a multivariate multimodal approach PLS was applied in an analysis of 5HT2A BP in conjunction with GM concentration and of 5HT2A BP in conjunction with 5HTT BP, respectively. In both analyses a latent variable associated with the PLS model was found to be highly correlated with age. Conclusion: Even though the integration of more modalities into a single analysis is in the very beginning, this thesis outlined some widespread future possibilities and challenges.
Series IMM-M.Sc.-2008-119
Original PDF ep08_119_web.pdf (3.53 MB)
Admin Creation date: 2008-11-18    Update date: 2008-11-18    Source: dtu    ID: 228763    Original MXD