Epsilon Open Archive

Abstract

Digital image analysis enables creating objective, fast, and reproducible analysis methods of objects or situations that can be imaged. This thesis contains theoretical work regarding distance transforms for images digitized in elongated grids. Such images are the result of many, mainly 3D, imaging devices. Local weights appropriate for different elongation factors in 2D, as well as in 3D, are presented. Methods adapted to elongated grids save time and computer memory compared to increasing the image size by interpolating to a cubic grid. A number of segmentation methods for images in specific applications are also included in the thesis. Distance information is used to segment individual pores in paper volume images. This opens the possibility to investigate how the pore network affects the paper quality. Stable and reliable segmentation methods for cell nuclei are necessary to enable studies of tumor morphology, as well as amounts of fluorescence marked substances in individual nuclei. Intensity, gradient magnitude, and shape information is combined in a method to segment cell nuclei in 2D fluorescence and 3D confocal microscopy images of tissue sections. Two match based segmentation methods are also presented. Three types of viral capsids are identified and described based on their radial intensity distribution in transmission electron micrographs of infected cells. This can be used to measure how a potential drug affects the relative amounts of the three capsids, and possibly, the viral maturation pathway. Proteins of a specific kind in transmission electron volume images of a protein solution are identified using a shape based match method. This method reduces the amount of visual inspection needed to identify proteins of interest in the images. Two representation schemes, developed in order to simplify the analysis of individual proteins in volume images of proteins in solution, are presented. One divides a protein into subparts based on the internal intensity distribution and shape. The other represents the protein by the maximum intensity curve connecting the centers of the subparts of the protein. These representations can serve as tools for collecting information about how flexible a protein in solution is and how it interacts with other proteins or substances. This information is valuable for the pharmaceutical industry, when developing new drugs.