- 152 views
- 97 downloads
Low Rank plus Sparse Decomposition of fMRI Data with Applications in Alzheimer’s Disease
-
- Author / Creator
- Fu, Fangfang
-
Functional brain connectivity plays an important role on understanding
how human brain functions and neuropsychological diseases such as Autism,
Attention-deficit hyperactivity disorder (ADHD), and Alzheimer’s disease (AD).
Functional Magnetic Resonance Imaging (fMRI) is one of the most powerful
techniques to construct functional brain connectivity. However, the presence
of potential outliers in fMRI BOLD signals might lead to unreliable results
in construction. Another challenge of most existing connectivity construction
methods is the results might not be stable. In this thesis, we propose a framework
which is able to provide robust and stable connectivity. In particular,
a low-rank plus sparse (L + S) matrix decomposition technique is adapted to
decompose the resting state fMRI BOLD signals, where the low-rank matrix
L recovers the essential common features from regions of interest (ROIs), and
the sparse matrix S catches the sparse individual variability and potential outliers.
Based on decomposition, various approaches can be applied to construct
functional brain connectivity, such as correlation, partial correlation, Graphical
Lasso and others. However, we propose to use the recently developed novel
sparse matrix estimation based on concentration inequality. Statistical test for
each connection is implemented for differentiating group difference. Through
bootstrapping afterwards, we verify whether performing low-rank plus sparse
matrix decomposition can achieve more stable and robust results.
We apply our method on Alzheimer’s Disease Neuroimaging Initiative (ADNI)
data, and compare the results with those based on original BOLD signals. We
discover that the methods for building connectivities based on low rank matrices behave better than based on original BOLD signals, in the sense that
the former can reveal and identify more significant ROI connections. We find
that the recently proposed concentration inequality based method performs
better overall compared with correlation, partial correlation, and Graphical
Lasso method. We also obtain the first ten most significant connections for
differentiating group differences for Alzheimer’s disease. Among them, the
left hippocampus region and the left cerebellum 7 region is the most significant
one, with p value smaller than 0.0005, which is consistent with existing
literatures’ findings. Through bootstrapping, we verify that performing low rank
plus sparse matrix decomposition can achieve more stable results for
constructing functional brain connectivities. -
- Subjects / Keywords
-
- Graduation date
- Fall 2018
-
- Type of Item
- Thesis
-
- Degree
- Master of Science
-
- License
- Permission is hereby granted to the University of Alberta Libraries to reproduce single copies of this thesis and to lend or sell such copies for private, scholarly or scientific research purposes only. Where the thesis is converted to, or otherwise made available in digital form, the University of Alberta will advise potential users of the thesis of these terms. The author reserves all other publication and other rights in association with the copyright in the thesis and, except as herein before provided, neither the thesis nor any substantial portion thereof may be printed or otherwise reproduced in any material form whatsoever without the author's prior written permission.