Segmentation of Lung Structures in CT
Chronic Obstructive Pulmonary Disease (COPD) or Smokers' Lungs is a major disease burden worldwide. Computed Tomography (CT) scans has been shown to be a viable solution for observing the effects and identifying the different subtypes of COPD, such as emphysema, which is the destruction of lung tissues, and bronchitis, the inflammation of the airway walls. The aim of this thesis is to develop algorithms for the segmentation of the lungs structures for the study of COPD using CT scans from lung cancer screening trials.
Segmentation is the process of finding regions in an image belonging to specific objects that are to be subjected for further processing, and is often the first step in image analysis. In this thesis, we have developed fully automated solutions for the segmentation of the lungs, pulmonary airways and vessels in CT. A unique approach that we have explored for pulmonary airways segmentation is the usage of machine learning techniques, where special algorithms were used to differentiate between airways and lung tissues via a given set of examples. Besides this, we also developed a segmentation technique that specializes in tree-like structures for extracting the pulmonary airways and vessels.
We have also conducted the Extraction of Airways from CT 2009 (EXACT'09), which is the first comparative study to evaluate the performance of different airway tree segmentation algorithms in a standardized manner. The comparison was performed with reference to an "ideal" airway tree constructed from the airway trees segmented by the different algorithms. A total of 15 algorithms were evaluated on a dataset consisting of various types of chest CT scans from eight medical institutes.
Chairman: Associate Professor Kim Steenstrup Pedersen, Department of Computer Science, Copenhagen University
Member 1: Associate Professor Matthew Brown, UCLA Radiological Sciences
Member 2: Principal Scientist Cristian Lorenz, Philips Research Europe, Hamburg
For an electronic copy of the thesis, please contact Dina Riis Johannessen, 3532 1423, email@example.com.