Segmentation of cell structures in fluorescence confocal microscopy images

Abstract

During the past several years, image segmentation techniques have been developed and extensively used in biomedical applications as an important tool to extract objects and boundaries of interest. In biological field, cytoskeleton analysis is a complicated problem and the analysing technique is still immature. Cytoskeleton plays an important role in normal cell activities, including motion and division, which make the cell cytoskeleton important to investigate. The objective of this project is to investigate and evaluate level set segmentation methods for segmentation of both cell nuclei and membrane segmentation of microfilament images captured by fluorescent confocal microscopy. Based on some background investigations, the active contour methodology has been selected as the fundamental method for image segmentation. This thesis presents the methods used and reports on the results achieved for cell and nuclei segmentation using the hybrid level-set method and cell membrane segmentation using the subjective surfaces model. In addition, some initial results of nuclei segmentation in 3-D case based on the hybrid method will be presented as well. Also included in this thesis are the method and the initial categorisation of microtubule images based on the multi-template method. At the end of the thesis, possible directions for potential future work are presented. It is envisaged that the segmentation tools produced by the project will make cell cytoskeleton data analysis much more convenient. In particular, the segmentation of cell membranes will help biologists to perform quantitative analysis of fluorescent confocal microscopy images by measuring the cell properties. With more useful information of cytoskeleton being provided, the work contained in this thesis has the potential to contribute to evaluation and prediction of the possibility of cell canceration.