The Library
Latent variable modelling of population neuroimaging and behavioural data
Tools
Liu, Zhangdaihong (2020) Latent variable modelling of population neuroimaging and behavioural data. PhD thesis, University of Warwick.
|
PDF
WRAP_Theses_Liu_2020.pdf - Submitted Version - Requires a PDF viewer. Download (43Mb) | Preview |
Official URL: http://webcat.warwick.ac.uk/record=b3467688~S15
Abstract
Neuroimaging has aroused much interest in recent years due to the growth of Magnetic Resonance Imaging (MRI) technology and data acquisition techniques. This has led to an increase in interest for work that links neuroscience to behavioural research using neuroimaging data to reveal the interplay between brain and behaviours. Latent variable models are popular tools to investigate such relationships, with many studies exploring links between functional MRI and various behavioural and demographic measures. However, a common challenge is the interpretability of the latent variable models, in particular, their applications to large datasets with thousands of variables. In this thesis, we first introduced the basic concepts in neuroimaging and the challenges faced when linking it to behaviours. Then, we introduced the background methods applied in the thesis including latent variable models, predictive models and some widely applied data processing techniques. The discussion focused on clarifying easily confused and misused concepts, the theory and application of some rare model extensions, and the demonstration of crossvalidation in chained latent variable models. Many of these notes, to our knowledge, have not been discussed elsewhere. One of the main focuses and contributions of this thesis is the proposal of a dimension reduction method, namely Supervised Dimension Reduction. It aims to improve the interpretation of latent variable models, especially in the application of chaining multiple models together. We applied Supervised Dimension Reduction together with other latent variable models to the Human Connectome Project and the UK Biobank project to study the relationships between neuroimaging and behavioural data. We revealed many interesting patterns between brain and behaviours. Moreover, we further clarified the interpretation of a commonly applied latent variable model, Canonical Correlation Analysis. In particular, the multi-view extension and their applications in brain-behaviour study. In the end, we attempted to use functional MRI to predict a specific behavioural measure: personality. However, no results turned out to be significant under the analysis pipeline we applied.
Item Type: | Thesis (PhD) | ||||
---|---|---|---|---|---|
Subjects: | B Philosophy. Psychology. Religion > BF Psychology Q Science > QA Mathematics |
||||
Library of Congress Subject Headings (LCSH): | Brain -- Imaging, Magnetic resonance imaging, Psychophysiology, Population psychology, Human behavior -- Mathematical models | ||||
Official Date: | April 2020 | ||||
Dates: |
|
||||
Institution: | University of Warwick | ||||
Theses Department: | Mathematics Institute | ||||
Thesis Type: | PhD | ||||
Publication Status: | Unpublished | ||||
Supervisor(s)/Advisor: | Nichols, Thomas E. ; Feng, Jianfeng | ||||
Sponsors: | China Scholarship Council ; University of Warwick. Mathematics for Real-World Systems Centre for Doctoral Training ; Alan Turing Institute ; Guarantors of Brain | ||||
Format of File: | |||||
Extent: | xxvii, 237 leaves : illustrations, charts | ||||
Language: | eng |
Request changes or add full text files to a record
Repository staff actions (login required)
View Item |
Downloads
Downloads per month over past year