Many practical applications in the field of medical image processing require valid and reliable segmentation of images. In this dissertation, we propose three different semi-automatic segmentation frameworks for 2D-upper torso medical images to construct 3D geometric model of the torso structures. In the first framework, an extended version of the Otsu&rsquo / s method for three level thresholding and a recursive connected component algorithm are combined. The segmentation process is accomplished by first using Extended Otsu&rsquo / s method and then labeling in each consecutive slice. Since there is no information about pixel positions in the outcome of Extended Otsu&rsquo / s method, we perform some processing after labeling to connect pixels belonging with the same tissue. In the second framework, Chan-Vese (CV) method, which is an example of active contour models, and a recursive connected component algorithm are used together. The segmentation process is achieved using CV method without egde information as stopping criteria. In the third and last framework, the combination of watershed transformation and K-means are used as the segmentation method. After segmentation operation, the labeling is performed for the determination of the medical structures. In addition, segmentation and labeling operation is realized for each consecutive slice in each framework. The results of each framework are compared quantitatively with manual segmentation results to evaluate their performances.

Na začátku diplomové práce je čtenář seznámen s procesem zpracování obrazu a v navazující části jsou popsáný a vysvětleny dnes nejpoužívanější algoritmy pro segmentaci obrazu. Na základě watershed transform je vytvořen segmentační operátor pro volně šiřitelný program Rapid Miner a v dokumentu je popsáno, jak proces vývoje probíhal. V poslední části práce jsou prezentovány segmentované obrazy a popsána úskalí takto implementované watershed transform metody.

This thesis explores two possible ways of detecting soft plaque present in the coronary arteries, using CTA imagery. The coronary arteries are vessels that supply oxidized blood to the cardiac muscle and are thus important for the proper functioning of heart. Cholesterol or reactive oxygen species from cigarette smoke and other toxins may get adhered to the walls of coronary arteries and trigger chronic inflammation that leads to formation of the soft plaque. When the soft plaque grows bigger in volume, it occludes the blood flow to the cardiac muscle and finally results in ischemic heart attack. Moreover, smaller plaque can easily rupture due to the blood flow in arteries and can result in complications such as stroke. Hence there is a need to detect the soft plaque using non-invasive or minimally invasive techniques. In CTA imagery, the cardiac muscle appears as a dark gray color, while the blood appears as dull white color and the the calcified plaque appears as bright white. The soft plaque has an intensity which falls between the intensity level of the blood and cardiac muscle, making it difficult to directly segment the soft plaque using standard segmentation methods. Soft plaque in its advanced stages forms a concavity in the blood lumen. A watershed based segmentation method was used to detect the presence of this concavity which in turn identifies the location of the soft plaque. For segmenting the soft plaque at its earlier stages, a novel segmentation technique was used. In this technique the surface is evolved based on a region-based energy calculated in the local neighborhood around each point on the evolving surface. This method seems to be superior to the watershed based segmentation method in detecting smaller plaque deposits.

This thesis describes a set of image analysis tools developed for the purpose of quantifying the distribution of chromatin in (light) microscope images of cell nuclei. The distribution or pattern of chromatin is influenced by both external and internal variations of the cell environment, including variations associated with the cell cycle, neoplasia, apoptosis, and malignancy associated changes (MACs). The quantitative characterisation of this pattern makes possible the prediction of the biological state of a cell, or the detection of subtle changes in a population of cells. This has important application to automated cancer screening. The majority of existing methods for quantifying chromatin distribution (texture) are based on the stochastic approach to defining texture. However, it is the premise of this thesis that the structural approach is more appropriate because pathologists use terms such as clumping, margination, granulation, condensation, and clearing to describe chromatin texture, and refer to the regions of condensed chromatin as granules, particles, and blobs. The key to the structural approach is the segmentation of the chromatin into its texture primitives. Unfortunately all of the chromatin segmentation algorithms published in the literature suffer from one or both of the following drawbacks: (i) a segmentation that is not consistent with a human's perception of blobs, particles, or granules; and (ii) the need to specify, a priori, one or more subjective operating parameters. The latter drawback limits the robustness of the algorithm to variations in illumination and staining quality. The structural model developed in this thesis is based on several novel low-, med-ium-, and high-level image analysis tools. These tools include: a class of non-linear self-dual filters, called folding induced self-dual filters, for filtering impulse noise; an algorithm, based on seeded region growing, for robustly segmenting chromatin; an improved seeded region growing algorithm that is independent of the order of pixel processing; a fast priority queue implementation suitable for implementing the watershed transform (special case of seeded region growing); the adjacency graph attribute co-occurrence matrix (AGACM) method for quantifying blob and mosaic patterns in the plane; a simple and fast algorithm for computing the exact Euclidean distance transform for the purpose of deriving contextual features (measurements) and constructing geometric adjacency graphs for disjoint connected components; a theoretical result establishing an equivalence between the distance transform of a binary image and the grey-scale erosion of its characteristic function by an elliptic poweroid structuring element; and a host of chromatin features that can be related to qualitative descriptions of chromatin distribution used by pathologists. In addition, this thesis demonstrates the application of this new structural model to automated cervical cancer screening. The results provide empirical evidence that it is possible to detect differences in the pattern of nuclear chromatin between samples of cells from a normal Papanicolaou-stained cervical smear and those from an abnormal smear. These differences are supportive of the existence of the MACs phenomenon. Moreover the results compare favourably with those reported in the literature for other stains developed specifically for automated cytometry. To the author's knowledge this is the first time, based on a sizable and uncontaminated data set, that MACs have been demonstrated in Papanicolaou stain. This is an important finding because the primary screening test for cervical cancer, the Papanicolaou test, is based on this stain.