研究期間：10108~10207;Accurate blood vessel segmentation is the most important process in computer vision and computer-aided diagnosis systems. However, the various diameters of targeted blood vessels, varied contrast agent concentration in the blood, and different circulation rates have made blood vessel segmentation the most complicated and difficult step in medical image processing. We examine a new approach called multiscale two-dimensional matched filters to segment blood vessels in retina images. Unlike previously published work, this new method uses multiplication to collect matched filtering responses under different scales. The results look better and promising. In this research project, we intend to extend this newly developed multiscale method to the field of three-dimensional (3D) blood vessel segmentation for CT angiography. In order to be able to study the new method's performance thoroughly, we will conduct a series of phantom studies to validate the segmentation accuracy. With the help of phantom studies, we hope we can incorporate the new method to our computer-aided diagnosis system to improve diagnostic accuracy for 3D blood vessel images.