Logo Logo
Hilfe
Kontakt
Switch language to English
Automatic approaches for microscopy imaging based on machine learning and spatial statistics
Automatic approaches for microscopy imaging based on machine learning and spatial statistics
One of the most frequent ways to interact with the surrounding environment occurs as a visual way. Hence imaging is a very common way in order to gain information and learn from the environment. Particularly in the field of cellular biology, imaging is applied in order to get an insight into the minute world of cellular complexes. As a result, in recent years many researches have focused on developing new suitable image processing approaches which have facilitates the extraction of meaningful quantitative information from image data sets. In spite of recent progress, but due to the huge data set of acquired images and the demand for increasing precision, digital image processing and statistical analysis are gaining more and more importance in this field. There are still limitations in bioimaging techniques that are preventing sophisticated optical methods from reaching their full potential. For instance, in the 3D Electron Microscopy(3DEM) process nearly all acquired images require manual postprocessing to enhance the performance, which should be substitute by an automatic and reliable approach (dealt in Part I). Furthermore, the algorithms to localize individual fluorophores in 3D super-resolution microscopy data are still in their initial phase (discussed in Part II). In general, biologists currently lack automated and high throughput methods for quantitative global analysis of 3D gene structures. This thesis focuses mainly on microscopy imaging approaches based on Machine Learning, statistical analysis and image processing in order to cope and improve the task of quantitative analysis of huge image data. The main task consists of building a novel paradigm for microscopy imaging processes which is able to work in an automatic, accurate and reliable way. The specific contributions of this thesis can be summarized as follows: • Substitution of the time-consuming, subjective and laborious task of manual post-picking in Cryo-EM process by a fully automatic particle post-picking routine based on Machine Learning methods (Part I). • Quality enhancement of the 3D reconstruction image due to the high performance of automatically post-picking steps (Part I). • Developing a full automatic tool for detecting subcellular objects in multichannel 3D Fluorescence images (Part II). • Extension of known colocalization analysis by using spatial statistics in order to investigate the surrounding point distribution and enabling to analyze the colocalization in combination with statistical significance (Part II). All introduced approaches are implemented and provided as toolboxes which are free available for research purposes., Unbekannt
Not available
Norousi, Ramin
2014
Englisch
Universitätsbibliothek der Ludwig-Maximilians-Universität München
Norousi, Ramin (2014): Automatic approaches for microscopy imaging based on machine learning and spatial statistics. Dissertation, LMU München: Fakultät für Mathematik, Informatik und Statistik
[thumbnail of Norousi_Ramin.pdf]
Vorschau
PDF
Norousi_Ramin.pdf

7MB

Abstract

One of the most frequent ways to interact with the surrounding environment occurs as a visual way. Hence imaging is a very common way in order to gain information and learn from the environment. Particularly in the field of cellular biology, imaging is applied in order to get an insight into the minute world of cellular complexes. As a result, in recent years many researches have focused on developing new suitable image processing approaches which have facilitates the extraction of meaningful quantitative information from image data sets. In spite of recent progress, but due to the huge data set of acquired images and the demand for increasing precision, digital image processing and statistical analysis are gaining more and more importance in this field. There are still limitations in bioimaging techniques that are preventing sophisticated optical methods from reaching their full potential. For instance, in the 3D Electron Microscopy(3DEM) process nearly all acquired images require manual postprocessing to enhance the performance, which should be substitute by an automatic and reliable approach (dealt in Part I). Furthermore, the algorithms to localize individual fluorophores in 3D super-resolution microscopy data are still in their initial phase (discussed in Part II). In general, biologists currently lack automated and high throughput methods for quantitative global analysis of 3D gene structures. This thesis focuses mainly on microscopy imaging approaches based on Machine Learning, statistical analysis and image processing in order to cope and improve the task of quantitative analysis of huge image data. The main task consists of building a novel paradigm for microscopy imaging processes which is able to work in an automatic, accurate and reliable way. The specific contributions of this thesis can be summarized as follows: • Substitution of the time-consuming, subjective and laborious task of manual post-picking in Cryo-EM process by a fully automatic particle post-picking routine based on Machine Learning methods (Part I). • Quality enhancement of the 3D reconstruction image due to the high performance of automatically post-picking steps (Part I). • Developing a full automatic tool for detecting subcellular objects in multichannel 3D Fluorescence images (Part II). • Extension of known colocalization analysis by using spatial statistics in order to investigate the surrounding point distribution and enabling to analyze the colocalization in combination with statistical significance (Part II). All introduced approaches are implemented and provided as toolboxes which are free available for research purposes.

Abstract