Parallelization of a data-driven independent component analysis to analyze large 3D-polarized light imaging data sets / vorgelegt von Dipl.-Ing. Giuseppe Teodoro Maria Tabbì aus Dortmund. Wuppertal, April 2016
Inhalt
- Kurzfassung
- Abstract
- Introduction
- Principles of 3D-Polarized Light Imaging
- Histological Preparation of Brain Tissue
- Polarimetric Measurement
- Jones Calculus applied on 3D-PLI
- Fourier Analysis of the 3D-PLI Signal
- Workflow
- Principles of the Independent Component Analysis
- Blind Source Separation
- Information Based Maximization
- Ambiguities of Independent Component Analysis
- Pre-processing
- Constrained Independent Component Analysis for 3D-PLI
- High Performance Computing
- System Configuration of JuDGE
- Parallel Programming Paradigms
- Parallel Programming Strategies
- Parallel Programming Models
- Development of an Adapted Constrained Independent Component Analysis on White and Gray Matter
- Parameter Optimization of the Adapted Constrained Independent Component Analysis
- A New Parallelization Concept for Adapted Constrained Independent Component Analysis
- Discussion
- Conclusion and Outlook
- Derivation of the 3D-PLI Signal
- Derivation of the Natural-gradient Version of Infomax
- Nongaussianity for Source Signals in Independent Component Analysis
- Basis vectors
- List of Figures
- List of Tables
- List of Algorithms
- Bibliography
- Acknowledgements