Measurement of Carotid Artery Wall Thickness and Diameters from Magnetic Resonance and Ultrasound Images

Abstract

One of the most common causes of cardiovascular diseases (CVD) is atherosclerosis which is the continuous build-up of fatty deposits (plaques) on the inner walls of the blood vessels (arteries). The incidence of ischemic strokes is highly associated with the rupture of atherosclerotic plaques in the common carotid artery (CCA). To assess carotid atherosclerosis, non-invasive imaging modalities such as magnetic resonance (MR) and ultrasound (US) imaging are preferred over other invasive methods due to their safer profile and ability to explore atherosclerosis in its early stages. It has been increasingly accepted that the wall thickness and diameter measurements of the CCA can serve as early indicators of CVD development. However, manual measurement of these quantities is tedious, error-prone and subjected to observer variability. Hence, there is a growing interest for the development of automated software systems for the measurement of wall thickness and diameters of the CCA from MR and US images. The development of such automated systems was the primary objective of this research. MR and US images are generally corrupted with noise which makes their interpretation and further processing difficult. This necessitates the task of denoising as a pre-processing stage which can improve the performance of image segmentation techniques. For this reason, a robust denoising filter (NLMLDCT ) has been proposed in this thesis to reduce the Rician noise in MR images by integrating discrete cosine transform into the conventional non-local maximum likelihood (NLML) method. Whereas, one of the widely accepted method named OBNLM is adopted in our work to reduce the speckle noise in US images. We have proposed novel algorithms for delineation of the carotid artery borders from MR and US images. The wall thickness of the CCA has been measured from T1- iiiweighted MR images using an active contour method combined with localized particle swarm optimization. This combination of global and localized segmentation strategy helped us to reduce the edge leaking problem to some extent. Due to better reliability, the diameters of the CCA have been measured from B-mode US images via a region-based approach. For this measurement, we have delineated the lumen-intima and media-adventitia borders of the CCA using a scale-space based strategy. This automated segmentation system has been further improved using an iterative spatial transformation based technique for handling curved vessels. Comprehensive statistical data analysis was performed to ensure the superior performance of the proposed techniques against the manual expert tracings.