Sherif Moursi, December 2007, "Semi-Automatic Snake Based Segmentation of Carotid Artery Ultrasound Images", Computer Science Department, University of Western Ontario, Canada.
M.Sc. Thesis Abstract
Carotid ultrasound imaging is one of the clinical diagnostic procedures that can be employed to detect plaque buildup at the carotid artery walls. It is an inexpensive and non-invasive procedure that has no known side effects. Yet, the acquired ultrasound images have poor quality and contain a lot of noise. Active contouring segmentation technique (also known as snakes or deformable model) is characterized by its robustness to both image noise and boundary gaps. Hence it is suitable to be used as a segmentation technique in noisy poor quality ultrasound images.
One of the major issues of active contouring method is its sensitivity to the initial contour that is provided by the user. Unless it is drawn close enough to the actual contour, it may lead to unsatisfactory results. Thus, most active contour algorithms require considerable user interaction to provide a good initial contour. This thesis presents a fast and efficient algorithm for extracting carotid artery lumens in ultrasound images. Our algorithm reduces the user interaction as the user is only required to place a seed point inside the region of interest. It utilizes our proposed rule-based initial contour, with carefully adjusting snake energies to refine this initial contour.
Sensitivity, precision rate, and overlap ratio have been used to measure the accuracy of the extracted contours. The results show that the extracted initial and final contours have a good overlap with contours that are manually segmented by an experienced clinician. Furthermore, our proposed initial contour generation scheme can be used as an independent module with any active contouring algorithm to produce accurate results.