Sadri, Neda (2018) Detection of thin structures in biomedical images. (PhD thesis), Kingston University, .
Abstract
Lung cancer is life-threatening and difficult to treat. According to the World Health Organization, lung cancer is categorised by uncontrolled cell growth in tissues of the lung and is the most common cancer with 1.59 million deaths worldwide in 2012. There are estimated to have been 1.8 million new cases of lung cancer in 2012. About 20% of lung cancer cases are not thought to be related to smoking. Vessel characteristics may change in association with tumors for example due to angiogenesis which is a fundamental component in the development of tumours. Radiological images including MRI and CT can detect lung tumours and surrounding vasculature. Manual detection of vessel-like structures is time-consuming. Thus, computer-assisted detection of vessel-like structures may help in tumour assessment. Cilia are membrane-bounded microtubles-based extensions of the centrosome that have different roles in mammalian development and adult physiology. Disorders of cilia or ciliopathies are associated with a number of genetic disorders such as sinus inversus. Cilia-like blood vessels are thin structures and assessment of number and length are considered important in the detection of disease. Manual detection of cilia is also difficult and time-consuming. This project divided into two applications of thin structures in biomedical images. First, the detection of the thin structures of cilia from microscopy images was performed used different techniques with aspects of linearity. Second, the best segmentation technique was created and developed to detect the cilia from microscopy images stablished and applies on the thin structure of vessels in CT scan images. The quantifications of both thin structures of cilia and vessels such as numbers and lengths were investigated. The aim of this thesis is to develop and apply techniques for detection of thin structures in medical images with particular reference to microscopy images of cilia and CT images of vessel-like structures in the vicinity of a lung tumour. A semi-automatic method was developed that combines mathematical morphological operations to enhance the thin objects combined with global thresholding, followed by user interaction methods to detect overlap and disconnected objects. The system was successfully applied to detect cilia from electron microscope images and to detect vessel-like structures in CT images of lung cancers. The techniques were applied to assess a number of features such as number, length and tortuosity. In a study of unilateral lung tumour image sets, a statistically significant difference was detected in the number of vessel-like structures in the region of lung tumours compared with the contralateral side with no tumour. Thus these methods may have application in detecting thin structures in microscope images as well as CT or other medical imaging modalities.
Actions (Repository Editors)
 |
Item Control Page |