Herold, Julia: A data mining approach for high-content fluorescence microscopy images of tissue samples. 2010
Inhalt
- Introduction
- Imaging for Protein Colocation Studies
- An Exploration System for Multivariate Fluorescence Tissue Images
- Existing Methods for the Exploration of Multivariate TIS Data
- Demands on an Object-Based Analysis Strategy for Non-Binary Multivariate Images
- Summary
- Supervised Learning-Based Object Detection in Tissue Micrographs
- Accuracy Assessment
- Human Synapse Detection Performance
- Requirements Posed on a Supervised Learning-Based Object Detection Strategy
- SVM Theory
- Linear Support Vector Machines
- Non Linear Support Vector Machines
- Parameter Selection
- Probabilistic Outputs for SVMs
- The i3S for Object Detection
- Image Preprocessing
- Training Set Generation
- i3S Training
- Computation of Object Positions
- Estimating a Constant Confidence Threshold
- Case study I: Brain Tissue
- The Influence of Training Set and Threshold Choice
- The Influence of the Labeling Strategy and Constant Thresholds in an Inter-Image Detection Setup
- i3S Performance in a Multi Protein Detection Setup
- Discussion
- Case study II: Pancreas Tissue
- Summary
- A Direct Visual Data Mining Tool for Three-Channel High-Content Micrographs
- Image Preprocessing
- Feature Extraction
- Interactive Visual Data Exploration and Cell Type Classification
- Results
- Discussion
- An Unsupervised Learning Approach for Data Mining d-dimensional Fluorescence Images
- Feature Calculation
- Cluster Analysis
- Online K-means
- Neural Gas
- Hierarchically Growing Hyperbolic Self Organizing Maps (H2SOM)
- Cluster Validation
- Selecting a Distance Function
- Visualizations of the Feature Domain for TIS Data Exploration
- Prototype Visualization
- Prototype Visualization via Combinatorial Intensity Profile Archetypes
- Cluster Visualization
- Visualizations in the Image Domain
- Statistical Comparison of Image Stacks
- Results
- Estimating the Appropriate Number of Clusters
- Stability of Cluster Results
- Analysis of Protein Colocation and Inter Image Correlation
- Comparison of Cluster Results of K-means, NG and H2SOM
- Analysis of the Whole Protein Set
- The Impact of Feature Calculation
- The Impact of Semantic Image Annotation
- Discussion
- Summary
- Statistical Synapse Distribution Analysis
- Ripley's K and O-Ring Statistic
- Null Models for Univariate and Bivariate Point Patterns
- Sample Application
- Discussion
- Conclusion and Outlook
- Appendix
- Bibliography