Options
2006
Report
Titel
Segmentation of human lung CT scans using hybrid method
Abstract
The purpose of this thesis work is to segment Human Lung CT Scans. Segmentation is the process of collecting image pixels or voxels based on the criteria of similarities and homogeneity in an image region. There are varying approaches to achieve this aim. As a part of this work combination of 'Region Growing' and 'Fuzzy Connected' methods are implemented as 'Hybrid Segmentation' method. These methods are based on the filters available in the 'Insight Segmentation and Registration Toolkit' (ITK). This thesis is divided in to three parts: as image acquisition plays an important role in the segmentation process, in part one of this thesis work history of different medical imaging technologies are explained along with the reasons why I was motivated to work in this direction. In the beginning of part two, basic structure and methods available for viewing medical images using Medical Imaging Platform (MIP, a platform developed at Fraunhofer Institute for the analysis of medical images) are explained. After that different steps involved in 'Pre-Processing' and 'Hybrid Segmentation' method of medical images are presented. In part three, the basic algorithm or pipeline is introduced, its implementation is explained, improved and finally tested with medical image data. In the testing phase, the algorithm is first tried with different datasets like 'water flask', 'Human Head CT Scan' and 'Human Lung CT Scan'. Finally, the results in each case of the algorithm or pipeline are presented and analysed to draw conclusions about the working procedure of this 'Hybrid Segmentation' method.
Beteiligt
Organisation
Fraunhofer-Institut für Graphische Datenverarbeitung -IGD-, Darmstadt
Verlagsort
Darmstadt