New Statistical Techniques for High-dimensional Neuroimaging Data Öffentlichkeit

Abstract

Abstract
In this dissertation, we focus on developing novel statistical methodology for high-
dimensional neuroimaging data (HND) to yield insights about the complex neural
processing characteristics associated with mental diseases and furthermore to provide
clinically relevant predictive information to aid with treatment selection and progno-
sis. Our proposed methods would extract new information content from neuroimaging
data, while coping with the many analytical challenges posed by these massive data
sets. Specifically, we propose three new statistical frameworks: (i) to predict disease
progression and therapeutic treatment response based on temporal trends in longi-
tudinal neuroimaging data, e.g. repeatedly collected over weeks or months; (ii) to
determine population-level brain networks revealed by functional (or structural) con-
nectivity properties within regions of interest (ROI) and between ROIs, while charac-
terizing whole-brain properties of these identified networks, such as the 'small world'
property; (iii) to analyze resting-state functional magnetic resonance imaging (fMRI)
data by constructing methodology to determine localized estimates of resting-state
brain activity and simultaneously yield functional connectivity information based on
spatial correlations in the data.

Table of Contents

Contents 1 Introduction 1 1.1 Overview . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1 1.2 Basic Knowledge of Human Brain . . . . . . . . . . . . . . . . . . . . 2 1.3 Neuroimaging Techniques . . . . . . . . . . . . . . . . . . . . . . . . 4 1.3.1 MRI and fMRI . . . . . . . . . . . . . . . . . . . . . . . . . . 5 1.3.2 PET . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 9 1.4 Preprocessing and Data Analysis Methods for Neuroimaging Data . . 10 1.4.1 Preprocessing Procedures . . . . . . . . . . . . . . . . . . . . 11 1.4.2 Statistical Modeling for Activation Studies . . . . . . . . . . . 12 1.4.3 Connectivity Analysis . . . . . . . . . . . . . . . . . . . . . . 14 1.4.4 Classication and Prediction . . . . . . . . . . . . . . . . . . . 20 1.5 Motivation Examples . . . . . . . . . . . . . . . . . . . . . . . . . . . 22 1.5.1 An fMRI Resting-state Study of Depression . . . . . . . . . . 22 1.5.2 A PET Study of Alzheimer's Disease . . . . . . . . . . . . . . 23 1.6 Proposed Research . . . . . . . . . . . . . . . . . . . . . . . . . . . . 23 2 Topic 1: A Novel Support Vector Classier for Longitudinal High- dimensional Data and Its Application to Neuroimaging Data 25 2.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 25 2.2 Methods . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 28 2.2.1 Classical Support Vector Classier . . . . . . . . . . . . . . . 28 2.2.2 Longitudinal Support Vector Classier - LSVC . . . . . . . . 30 2.2.3 Nonlinear Kernel Functions . . . . . . . . . . . . . . . . . . . 33 2.2.4 Feature Selection and Parameter Tuning . . . . . . . . . . . . 34 2.3 Results . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 35 2.3.1 Simulation study . . . . . . . . . . . . . . . . . . . . . . . . . 35 2.3.2 Data Example . . . . . . . . . . . . . . . . . . . . . . . . . . . 38 2.4 Discussion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 40 2.5 Appendix: Proof of Convexity of Objective function w.r.t. and . 41 3 Topic 2: Bayesian Hierarchical Model for Comprehensive Brain Con- nectivity Analysis 44 3.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 44 3.2 Motivating Example . . . . . . . . . . . . . . . . . . . . . . . . . . . 47 3.3 Methods . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 49 3.3.1 Bayesian Hierarchical Model . . . . . . . . . . . . . . . . . . . 49 3.3.2 Estimation and posterior inference . . . . . . . . . . . . . . . 54 3.4 Results . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 56 3.4.1 MDD Study Using Resting-state fMRI Data . . . . . . . . . . 56 3.4.2 Simulation Study . . . . . . . . . . . . . . . . . . . . . . . . . 62 3.5 Discussion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 65 4 Topic 3: An Unied Bayesian Framework for Resting-state fMRI Data Analysis: Jointly Modeling Frequency Activity and Connec- tivity 67 4.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 67 4.2 Motivating Example . . . . . . . . . . . . . . . . . . . . . . . . . . . 69 4.3 Method . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 70 4.4 Results . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 76 4.4.1 Voxel-level results . . . . . . . . . . . . . . . . . . . . . . . . . 76 4.4.2 Regional results . . . . . . . . . . . . . . . . . . . . . . . . . . 77 4.4.3 Regional connectivity networks . . . . . . . . . . . . . . . . . 80 4.5 Discussion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 84 5 Summary and Future Work 85

About this Dissertation

Rights statement

• Permission granted by the author to include this thesis or dissertation in this repository. All rights reserved by the author. Please contact the author for information regarding the reproduction and use of this thesis or dissertation.

School	Laney Graduate School
Department	Biostatistics
Degree	PhD
Submission	Dissertation
Language	English
Research Field	Biology, Biostatistics
Stichwort	Resting-state fMRI Brain connectivity Machine Learning Neuroimaging statistics High-dimensional
Committee Chair / Thesis Advisor	Bowman, F Dubois, Emory University
Committee Members	Waller, Lance, Emory University Guo, Ying, Emory University Hu, Xiaoping P, Emory University

Zuletzt geändert

Primary PDF

Thumbnail	Title	Date Uploaded	Actions
	New Statistical Techniques for High-dimensional Neuroimaging Data ()	2018-08-28 10:37:10 -0400	Press to Select an action Download

Supplemental Files

Thumbnail	Title	Date Uploaded	Actions