Wüllems, Karsten: Data-Driven Approaches to Exploratory Visual Analysis of Mass Spectrometry Imaging Data. 2021
Inhalt
- Titlepage
- Zusammenfassung
- Abstract
- Acknowledgment
- Contributions
- Contents
- 1 Introduction
- 1.1 Mass Spectrometry Imaging
- 1.1.1 Formal Definition of a Mass Spectrometry Imaging Data Cube
- 1.1.2 Matrix-Assisted Laser Desorption/Ionization
- 1.1.3 Time of Flight
- 1.1.4 Orbitrap
- 1.2 Analysis Approaches in Mass Spectrometry Imaging
- 1.3 Unsupervised Data-Driven Analysis
- 1.4 Visualization of Mass Spectrometry Imaging Data
- 1.5 Research Objectives
- 2 Data
- 3 Pre-Processing
- 3.1 Motivation
- 3.1.1 Dimension Reduction
- 3.1.2 High Dimensional Data
- 3.1.3 Using Embeddings to Explore Regions of Interest
- 3.2 Pre-Processing Pipeline
- 3.2.1 Alignment & Normalization
- 3.2.2 Reformatting
- 3.2.3 Matrix and Artifacts Detection and Reduction
- 3.2.4 Peak Picking and Deisotoping
- 3.2.5 Further Modules
- 3.2.6 Application on Real Data
- 3.3 Interactive Visual Exploration of Dimension Reduction in Mass Spectrometry Imaging
- 3.4 Summary and Contributions
- 3.5 Improvements and Future Research
- 4 Co-Localization Analysis in the Spatial Domain
- 4.1 Motivation
- 4.2 Quantification of Co-Localization of Molecular Distributions
- 4.2.1 Image Pattern Regularity
- 4.2.2 Mass Channel Image Features and Representations
- 4.2.3 Image Scale-Space Representations
- 4.2.4 Evaluation of Pipeline Setups
- 4.2.5 SoRC Score
- 4.2.6 Similarity Functions
- 4.2.7 Clustering Methods
- 4.2.8 Application on Real Data
- 4.2.9 Improvements and Future Research
- 4.3 Community Detection on m/z-Image Similarity Graphs
- 4.3.1 Building the m/z-Image Similarity Graph
- 4.3.2 Interactive Visual Exploration of m/z-Image Similarity Graphs
- 4.3.3 Improvements and Future Research
- 4.4 Approximation of Regions of Interest in the Spatial Domain
- 4.5 Summary and Contributions
- 5 Comparative Molecular Composition Analysis in the Spectral Domain
- 5.1 Motivation
- 5.2 Hierarchical Hyperbolic Self Organizing Maps
- 5.3 Segmentation Maps
- 5.3.1 H2SOM Projection
- 5.3.2 Ring-wise Position Optimization of the H2SOM Grid Projection
- 5.3.3 Projection of other Clustering Methods
- 5.3.4 Application on Real Data
- 5.4 Interactive Visual Exploration of Molecular Composition based Segmentation Maps
- 5.5 Summary and Contributions
- 5.6 Improvements and Future Research
- 6 Combining the Spatial and Spectral Domain for an Interactive and Responsive Analysis
- 6.1 Motivation
- 6.2 Interactive Visual Exploration of Molecular Composition Similarity through Spatial Browsing and Pseudocoloring
- 6.2.1 Efficient Implementation of QUIMBI
- 6.2.2 Advantages and Limitations
- 6.2.3 Application on Real Data
- 6.3 Summary and Contributions
- 6.4 Improvements and Future Research
- 7 Conclusion
- Bibliography
- Symbols
- A Appendix
- Colophon
- Declaration
- Declaration