Please use this identifier to cite or link to this item: http://bura.brunel.ac.uk/handle/2438/15018
Title: Consensus clustering framework for analysing fMRI datasets.
Authors: Liu, Chao
Advisors: Nandi, A
Meng, H
Keywords: Data-driven neuroimaging data analysis;Bi-Copam;Machine Learning;Affective processing;Scalable
Issue Date: 2017
Publisher: Brunel University London
Abstract: Neuroimaging of humans has gained a position of status within neuroscience. Modern functional magnetic resonance imaging (fMRI) technique provides neuroscientists with a powerful tool to depict the complex architecture of human brains. fMRI generates large amount of data and many analysis methods have been proposed to extract useful information from the data. Clustering technique has been one of the most popular data-driven techniques to study brain functional connectivity, which excels when traditional model-based approaches are difficult to implement. However, the reliability and consistency of many findings are jeopardised by too many analysis methods, parameters, and sometimes too few samples used. In this thesis, a consensus clustering analysis framework for analysing fMRI data has been developed, aiming at overcoming the clustering algorithm selection problem as well as reliability issues in neuroimaging. The framework is able to identify groups of voxels representing brain regions that consistently exhibiting correlated BOLD activities across many experimental conditions by integrating clustering results from multiple clustering algorithms and various parameters such as the number of clusters š¾. In the framework, the individual clustering result generation is aided by high performance grid computing technique to reduce the overall computational time. The integration of clustering results is implemented by a technique named binarisation of consensus partition matrix (Bi-CoPaM) adapted and enhanced for fMRI data analysis. The whole framework has been validated and is robust to participantsā€™ individual variability, yielding most complete and reproducible clusters compared to the traditional single clustering approach. This framework has been applied to two real fMRI studies that investigate brain responses to listening to the emotional music with different preferences. In the first fMRI study, three brain structures related to visual, reward, and auditory processing are found to have intrinsic temporal patterns of coherent neuroactivity during affective processing, which is one of the few data-driven studies that have observed. In the second study, different levels of engagement, i.e. intentional to unintentional, with music have unique effects on the auditory- limbic connectivity when listening to music, which has not been investigated and understood well in euro science of music field. We believe the work in this thesis has demonstrated an effective and competent approach to address the reliability and consistency concerns in fMRI data analysis.
Description: This thesis was submitted for the award of Doctor of Philosophy and was awarded by Brunel University London
URI: http://bura.brunel.ac.uk/handle/2438/15018
Appears in Collections:Electronic and Computer Engineering
Dept of Electronic and Computer Engineering Theses

Files in This Item:
File Description SizeFormat 
FulltextThesis.pdf File embargoed until 14/02/201813.57 MBAdobe PDFView/Open


Items in BURA are protected by copyright, with all rights reserved, unless otherwise indicated.