DOCSLIB.ORG

A Riemannian Framework for Shape Analysis of Subcortical Brain Structures

Total Page：16

File Type：pdf, Size：1020Kb

Download full-text PDF Read full-text

Load more

A Riemannian Framework for Shape Analysis of Subcortical Brain Structures A dissertation presented to the faculty of the Russ College of Engineering and Technology of Ohio University In partial fulfillment of the requirements for the degree Doctor of Philosophy Shuisheng Xie August 2013 © 2013 Shuisheng Xie. All Rights Reserved. 2 This dissertation titled A Riemannian Framework for Shape Analysis of Subcortical Brain Structures by SHUISHENG XIE has been approved for the School of Electrical Engineering and Computer Science and the Russ College of Engineering and Technology by Jundong Liu Associate Professor of Electrical Engineering and Computer Science Dennis Irwin Dean, Russ College of Engineering and Technology 3 Abstract XIE, SHUISHENG, Ph.D., August 2013, Electrical Engineering and Computer Science A Riemannian Framework for Shape Analysis of Subcortical Brain Structures (106 pp.) Director of Dissertation: Jundong Liu Measuring the volumetric and morphological changes of brain structures in MR images can provide a reliable basis for diagnosing and studying Alzheimer’s Disease (AD) and Mild Cognitive Impairment (MCI). Currently, the prevailing morphometric analysis frameworks are voxel-based morphometry (VBM), deformation-based morphometry (DBM) and tensor-based morphometry (TBM), which all involve a low-order spatial normalization procedure to project the computed voxel tissue map or motion field into a standard space. Shape and deformation information in local areas tend to be smoothed or even wiped out during the averaging process. In order to address the drawbacks of VBM/DBM/TBM, a novel local shape-based morphometry (SBM) framework on individual structures is developed in this dissertation, to measure the difference among a collection of anatomical images for individual subjects. SBM consists of three components: (1) a shape representation; (2) shape matching and comparison under Riemannian shape spaces; and (3) shape classifications and group analysis based on a well defined shape distance metric. Two different sets of solutions have been developed for the three components. The first approach relies on an information rich shape representation based on skeleton (IRS), and the dissimilarity between two shapes is defined as the geodesic distance connecting their projections on a shape manifold. The second model is based on a Riemannian meridian shape (RMS) representation. Stemmed from spectral graph theory, RMS possesses the merit of capturing the salient structural properties along the direction of maximal shape variation. Group clustering, based on pair-wise shape distances, is carried out through advanced manifold learning techniques. Experiments with synthetic shapes 4 and subcortical brain structures demonstrate the effectiveness of our framework in calculating the distances among different shapes and serving as a potential anatomical biomarker for neurodegenerative diseases, including AD. 5 To my parents. 6 Acknowledgments This dissertation would never have come to fruition without the supports from many professors, group mates, classmates and friends, and I would like to take this opportunity to express my sincere appreciation of their efforts. My deepest gratitude is to my advisor, Dr. Jundong Liu. His constant guidance, patience and support opened the door and guided me through a very rigorous yet enjoyable Ph.D. research training experience. His enormous enthusiasm and passion for work will always be an inspiration to me for many years to come. I would like to thank Dr. Chelberg for the helps he provided over the years. His solid background and deep understanding of the technology constantly enlighten my way to tackle the research problems I was facing. I really appreciate Dr. Marling’s encouragements to develop myself into an independent, mature and well-rounded researcher. Thanks to Dr. Drews. His suggestions to improve my research and dissertation are very valuable and insightful. I am grateful to the external committee members, Drs. Melkonian, McCall, Xu and Lin. It is really kind of them to spend their precious time serving on my committee, evaluating my research, and providing comments to improve my dissertation. Thanks to my group mates, Huihui Xu, Bibo Shi, David Days, Kevin H. Hobbs, Zhe Yu, Xin Qian, Pin Zhang and Yani Chen for the enjoyable discussions and collaborations over many interesting research topics. I am also indebted to many professors and friends at OU who have helped me over the years. I feel thankful to Dr. Shen. She has been a mentor for me in many aspects of my life. Thanks to Dr. Aizicovici and other professors in the Math department. Their kindness and considerateness towards the students are really admirable. 7 I need to thank the School of Electrical Engineering and Computer Science for the well designed program, as well as the financial support provided over the years. Without them, I would not have been able to complete my study and research. Last but not least, I would like to say thank you to my parents and other family members. Thanks for their understanding and supports. 8 Table of Contents Page Abstract.........................................3 Dedication........................................5 Acknowledgments....................................6 List of Tables...................................... 10 List of Figures...................................... 11 1 Introduction..................................... 13 2 Technical Background............................... 20 2.1 Riemannian Shape Space........................... 20 2.1.1 Basic Concepts............................ 20 2.1.2 Geodesics on Spheres........................ 22 2.1.3 Riemannian Statistics......................... 24 2.1.4 Some Recent Work of Shape Analysis on Manifolds........ 26 2.2 Skeletons for Shape Representation..................... 28 2.2.1 Curve Skeleton............................ 29 2.2.2 M-Rep and CM-Rep ......................... 33 2.3 Mesh Generation............................... 34 2.4 Spectral Graph Theory............................ 39 2.4.1 Graph Laplacian........................... 39 2.4.2 Laplacian Eigenmaps Embedding.................. 42 3 IRS Model for Shape-Based Morphometry..................... 44 3.1 Parameterization and Sphere Mapping.................... 45 3.2 IRS for 2D Shapes.............................. 46 3.3 IRS for 3D Shapes.............................. 47 3.4 Experimental Results............................. 47 3.5 Discussion................................... 53 4 Riemannian Shape Analysis based on Meridian Curves.............. 57 4.1 Related hippocampal structure analysis work................ 58 4.2 Mesh Generation............................... 61 4.3 Fiedler value and Fiedler vector....................... 63 4.4 Meridians Extraction based on Fiedler Vector................ 66 4.4.1 Meridians Extraction......................... 67 9 4.5 Riemannian Shape Space based on Meridians................ 71 4.6 Distances between RMSs........................... 74 4.7 Shape Clustering based on Inter-subject Distances.............. 75 4.7.1 Manifold Learning for RMS Dimensional Reduction........ 75 4.7.2 Shape Clustering........................... 77 4.8 Experimental Results............................. 78 4.8.1 Synthetic Examples.......................... 78 4.8.2 Subcortical structures from brain MR images............ 80 4.8.3 Clustering of AD Patients based on Hippocampi Shapes...... 84 4.8.4 Comparison with State-of-the-art Solutions............. 86 4.9 Discussion................................... 91 5 Summary and Future Work............................. 93 5.1 Summary................................... 93 5.2 Future Work.................................. 94 References........................................ 96 10 List of Tables Table Page 3.1 Pairwise shape distances using IRS model..................... 51 3.2 Pairwise shape distances using Joshi’s model [1]................. 52 4.1 Pairwise shape distances between shapes A-E .................. 79 4.2 Pairwise shape distances between the first 15 structures.............. 82 11 List of Figures Figure Page 1.1 Amyloid plaques and neurofibrillary tangles [2].................. 13 1.2 Brain atrophy in Alzheimer’s Disease (AD) [3].................. 14 1.3 Dynamic biomarkers of the Alzheimer’s pathological cascade [4]........ 15 1.4 Illustration of the VBM analysis procedure. [3].................. 17 1.5 Hippocampus as a subcortical component of human brains [5].......... 18 2.1 Tangent space, tangent vector and the curve.................... 21 2.2 Row-wise geodesic paths between the pair of curves shown to the left [1].... 28 2.3 Level curves and the curvilinear coordinate system in [6]............. 29 2.4 Skeleton of a 2D object............................... 30 2.5 3D objects and their corresponding surface skeletons [7]............. 31 2.6 Examples of curve-skeletons of different 3D objects [8]............. 31 2.7 Skeletonization results using the algorithm in [9]................. 33 2.8 The m-rep in [10]: Medial atom with a cross section of the boundary surface it implies (left). An m-rep model of a hippocampus and its boundary surface (right)........................................ 34 2.9 Quadtree decomposition [11]........................... 35 2.10 The advancing front triangulation method [11].................. 35 2.11 The empty circle property in Delaunay triangulation method [11]........ 36 2.12 Diagonal swapping in the Lawson’s Algorithm.................. 37 2.13 Bowyer-Watson Algorithm’s “void boundary” procedure............
Recommended publications
  • A Semi-Automated Workflow Solution for Multimodal Neuroimaging

    A Semi-Automated Workflow Solution for Multimodal Neuroimaging

    Brain Informatics (2016) 3:1–15 DOI 10.1007/s40708-015-0026-y A semi-automated workflow solution for multimodal neuroimaging: application to patients with traumatic brain injury Koon-Pong Wong . Marvin Bergsneider . Thomas C. Glenn . Vladimir Kepe . Jorge R. Barrio . David A. Hovda . Paul M. Vespa . Sung-Cheng Huang Received: 2 November 2015 / Accepted: 16 November 2015 / Published online: 1 December 2015 Ó The Author(s) 2015. This article is published with open access at Springerlink.com Abstract Traumatic brain injury (TBI) is a major cause 15O-water (0–10 min) and 18F-FDDNP (0–6 min) for of mortality and morbidity, placing a significant financial measuring cerebral blood flow in patients with TBI. burden on the healthcare system worldwide. Non-invasive neuroimaging technologies have been playing a pivotal Keywords Traumatic brain injury (TBI) Á Cerebral blood role in the study of TBI, providing important information flow (CBF) Á Magnetic resonance imaging (MRI) Á for surgical planning and patient management. Advances in 15O-water Á 18F-FDDNP Á Positron emission tomography understanding the basic mechanisms and pathophysiology (PET) Á Spatial normalization of the brain following TBI are hindered by a lack of reli- able image analysis methods for accurate quantitative assessment of TBI-induced structural and pathophysiolog- 1 Introduction ical changes seen on anatomical and functional images obtained from multiple imaging modalities. Conventional Traumatic brain injury (TBI) is an important public health region-of-interest (ROI) analysis based on manual labeling and socio-economic problem throughout the world. It is of brain regions is time-consuming and the results could be one of the most common causes of death and long-term inconsistent within and among investigators.
  • Spatial Normalization of Diffusion Tensor MRI Using Multiple Channels

    Spatial Normalization of Diffusion Tensor MRI Using Multiple Channels

    NeuroImage 20 (2003) 1995–2009 www.elsevier.com/locate/ynimg Spatial normalization of diffusion tensor MRI using multiple channels Hae-Jeong Park,a,b Marek Kubicki,a,b Martha E. Shenton,a,b Alexandre Guimond,c Robert W. McCarley,a Stephan E. Maier,d Ron Kikinis,b Ferenc A. Jolesz,d and Carl-Fredrik Westinb,* a Clinical Neuroscience Division, Laboratory of Neuroscience, Boston VA Health Care System-Brockton Division, Department of Psychiatry, Harvard Medical School, Boston, MA 02115, USA. b Surgical Planning Laboratory, MRI Division, Department of Radiology, Brigham and Women’s Hospital, Harvard Medical School, Boston, MA 02115, USA. c Center for Neurological imaging, MRI Division, Department of Radiology, Brigham and Women’s Hospital, Harvard Medical School, Boston, MA 02115, USA. d Department of Radiology, Brigham and Women’s Hospital, Harvard Medical School, Boston, MA 02115, USA. Received 10 March 2003; revised 30 July 2003; accepted 5 August 2003 Abstract Diffusion Tensor MRI (DT-MRI) can provide important in vivo information for the detection of brain abnormalities in diseases characterized by compromised neural connectivity. To quantify diffusion tensor abnormalities based on voxel-based statistical analysis, spatial normalization is required to minimize the anatomical variability between studied brain structures. In this article, we used a multiple input channel registration algorithm based on a demons algorithm and evaluated the spatial normalization of diffusion tensor image in terms of the input information used for registration. Registration was performed on 16 DT-MRI data sets using different combinations of the channels, including a channel of T2-weighted intensity, a channel of the fractional anisotropy, a channel of the difference of the first and second eigenvalues, two channels of the fractional anisotropy and the trace of tensor, three channels of the eigenvalues of the tensor, and the six channel tensor components.
  • Handbook of Functional MRI Data Analysis

    Handbook of Functional MRI Data Analysis

    Handbook of Functional MRI Data Analysis Functionalmagnetic resonance imaging (fMRI) has become the most popularmethod for imaging brain function. Handbook of Functional MRI Data Analysis provides a comprehensive and practical introduction to the methods used for fMRI data analysis. Using minimal jargon, this book explains the concepts behind processing fMRI data, focusing on the techniques that are most commonly used in the field. This book provides background about the methods employed by common data analysis packages including FSL, SPM, and AFNI. Some of the newest cutting-edge techniques, including pattern classification analysis, connectivity modeling, and resting state network analysis, are also discussed. Readers of this book, whether newcomers to the field or experienced researchers, will obtain a deep and effective knowledge of how to employ fMRI analysis to ask scientific questions and become more sophisticated users of fMRI analysis software. Dr. Russell A. Poldrack is the director of the Imaging Research Center and professor of Psychology and Neurobiology at the University of Texas at Austin. He has published more than 100 articles in the field of cognitive neuroscience, in journals including Science, Nature, Neuron, Nature Neuroscience, and PNAS. He is well known for his writings on how neuroimaging can be used to make inferences about psychologicalfunction, as wellasfor his research using fMRI and other imaging techniques to understand the brain systems that support learning and memory, decision making, and executive function. Dr. Jeanette A. Mumford is a research assistant professor in the Department of Psychology at the University of Texas at Austin. Trained in biostatistics, her research has focused on the development and characterization of new methods for statistical modeling and analysis of fMRI data.
  • Functional MRI User's Guide

    Functional MRI User's Guide

    Functional MRI User's Guide Michael A. Yassa ● The Division of Psychiatric Neuroimaging ● ● Department of Psychiatry and Behavioral Sciences ● ● The Johns Hopkins School of Medicine ● ● Baltimore, MD ● 1 Document written in OpenOffice.org Writer 2.0 by Sun Microsystems Publication date: June 2005 (1st edition) Online versions available at http://pni.med.jhu.edu/intranet /fmriguide/ Acknowledgments: This document relies heavily on expertise and advice from the following individuals and/or groups: John Ashburner, Karl Friston, and Will Penny (FIL-UCL: London), Kalina Christoff (UBC: Canada), Matthew Brett (MRC-CBU: Cambridge), and Tom Nichols (SPH-UMichigan, Ann Arbor). Some portions of this document are adapted or copied verbatim from other sources, and are referenced as such. Supplemental Reading: Frackowiak RS, Friston K, Frith C, Dolan RJ, Price CJ, Zeki S, Ashburner J, & Perchey G (2004). Human Brain Function, 2nd edition, Elsevier Academic Press, San Diego, CA. Huettel SA, Song AW, McCarthy, G. (2004) Functional Magnetic Resonance Imaging. Sinaur Associates, Sunderland, MA. 2 Table of Contents Magnetic Resonance Physics..............................................................................6 How the MR Signal is Generated.............................................................................6 The BOLD Contrast Mechanism..............................................................................8 Hemodynamic Modeling.........................................................................................10 Signal and Noise
  • Manual Computational Anatomy Toolbox - CAT12

    Manual Computational Anatomy Toolbox - CAT12

    Manual Computational Anatomy Toolbox - CAT12 Quick Start Guide 2 Version Information 4 Introduction and Overview 10 Getting Started 10 Download and Installation 10 Starting the Toolbox 11 Basic VBM Analysis (Overview) 11 Overview of CAT12 Processing 14 CAT12 Major Processing Steps 14 CAT12 Processing Steps in Detail 15 Basic VBM Analysis (detailed description) 17 Preprocessing Data 17 First Module: Segment Data 17 Second Module: Display slices (optionally) 18 Third Module: Estimate Total Intracranial Volume (TIV) 19 Fourth Module: Check sample 19 Fifth Module: Smooth 21 Building the Statistical Model 22 Two-sample T-Test 23 Full Factorial Model (for a 2x2 Anova) 24 Multiple Regression (Linear) 25 Multiple Regression (Polynomial) 26 Full Factorial Model (Interaction) 27 Full Factorial Model (Polynomial Interaction) 28 Estimating the Statistical Model 29 1 Checking for Design Orthogonality 29 Defining Contrasts 31 Special Cases 36 CAT12 for longitudinal data 36 Optional Change of Parameters for Preprocessing 39 Preprocessing of Longitudinal Data 39 Longitudinal Data in One Group 40 Longitudinal Data in Two Groups 42 Longitudinal Data in Two Groups with interaction of covariate by group 44 Adapting the CAT12 workflow for populations such as children 48 Customized Tissue Probability Maps 48 Customized Dartel- or Shooting-template 49 Other variants of computational morphometry 53 Deformation-based morphometry (DBM) 53 Surface-based morphometry (SBM) 54 Region of interest (ROI) analysis 57 Additional information on native, normalized and modulated volumes 58 Naming convention of output files 60 Calling CAT from the UNIX command line 62 Technical information 63 CAT12 Citation 66 References 68 2 Quick Start Guide VBM data ● Segment data using defaults (use Segment Longitudinal Data for longitudinal data).