Bioimage Informatics in the Big Data Era: Algorithms for High-Dimensional Spectral, Volumetric, and Temporal Image Processing translation missing: zh.hyrax.visibility.files_restricted.text

Rossetti, Blair J. (Summer 2019)

Permanent URL: https://etd.library.emory.edu/concern/etds/tq57ns13v?locale=zh
Published

Abstract

Big data is revealing new challenges in the area of bioimage informatics--the study of systems and methods for handling, processing, and visualizing biological images. Biological events are inherently multidimensional, and they occur at different timescales, frequencies, and resolutions. Yet, for many years, the limitations of digital imaging technologies have prevented researchers from gaining a holistic view of their problem. The goal of modern bioimage informatics is to leverage knowledge in big data, computer vision, and machine learning to efficiently and effectively process high-dimensional data sets. Although many open-source algorithms are available for processing bioimages, advanced numerical methods are still needed for handling big spectral, volumetric, and temporal data. In this work, we discuss novel contributions to the areas of spectral unmixing, volumetric reconstruction, and video-based tracking. Specifically, we present (1) a robust algorithm for unmixing large numbers of fluorescent labels from contaminated spectral micrographs; (2) an offline/online multiresolution workflow for the reconstruction of three-dimensional subvolumes of interest from serial gigapixel whole slide images; and (3) an automated analysis pipeline and graphical interface to aid in situ tracking of insect movement that uses low-cost, human-readable tags.

Table of Contents

1 Introduction - 1

1.1 Big Data and the Big Data Era - 2

1.2 Bioimage Informatics - 4

1.3 Contributions of Work - 6 

1.4 Outline - 7

2 Spectral Data - 8 

2.1 Spectral Imaging - 10

2.1.1 Applications in Microscopy - 16

2.2 Spectral Mixing and Unmixing - 20

2.2.1 Linear Mixing Model (LMM) - 20

2.2.2 Linear Unmixing - 22 

2.3 Spectral Mixing and Unmixing of Contaminated Micrographs - 30

2.3.1 Affine Mixing Model (AMM) - 31 

2.3.2 Semi-blind Sparse Affine Spectral Unmixing (SSASU) - 32

2.4 Comparison of Methods - 37

2.4.1 Endmember Estimation - 38

2.4.2 Spectral Unmixing - 40 

2.5 Discussion - 43 

3 Volumetric Data - 47 

3.1 Volumetric Imaging - 50 

3.2 Registration - 55

3.2.1 Intensity-Based Registration - 56

3.2.2 Feature-Based Registration - 57

3.2.3 Parametric Transformation - 58

3.2.4 Nonparametric Transformation - 61

3.3 Reconstruction - 61

3.3.1 Extensions to Gigapixel Volumetric Data - 63

3.4 Subvolume Registration from Gigapixel Serial Sections - 64

3.4.1 Registration - 65

3.4.2 Reconstruction by Composition of Scaled Transforms - 70

3.4.3 Evaluation - 72

3.5 Discussion - 77

4 Temporal Data 78

4.1 Temporal Imaging - 80

4.2 Object Tracking - 83

4.2.1 Applications in Ecology - 85

4.3 Graphical Insect Tracking Environment - 86

4.3.1 Video Preprocessing Module - 89

4.3.2 Tag Detection Module - 91

4.3.3 Digit Recognition Module - 93

4.3.4 Track Assembly Module - 94

4.3.5 Graphical Editor - 95

4.3.6 Evaluation - 96

4.4 Discussion. - 97

5 Conclusion - 99 

Appendix A Spectral - 102 

A.1 Sample Preparation - 102

A.2 Imaging and Preprocessing - 103

Appendix B Temporal - 104 

B.1 Experimental Setup - 104

Bibliography - 107 

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
Department
Degree
Submission
Language
  • English
Research field
关键词
Committee Chair / Thesis Advisor
Committee Members
最新修改 No preview

Primary PDF

Supplemental Files