Our goal of the proposed research is to develop a highly accurate and reproducible method to assess the severity of stenotic lesions in coronary arteriograms as well as in digital subtraction angiographic (DSA) images. The geometrical and functional parameters of stenotic lesions will be determined from opacified vessel images by using an iterative deconvolution technique and cross-correlation of the vessel contrast distributions along the vessel segment, i.e., the """"""""distance-density"""""""" curves. We believe that our method will provide an accurate anatomical assessment of atherosclerosis regression or progression in stenotic lesions as well as accurate dimensions of normal vessels, with which a reliable blood flow rate from angiograms can be determined for evaluation of hemodynamic characteristics of stenotic lesions. Specifically, we plan to develop methods for quantitative analysis of stenotic lesions using an iterative deconvolution technique on coronary arteriograms, by incorporating a sector search method for tracking of vessel segments, correction of nonuniform background trend using a 2D surface fitting technique, and analysis of symmetry in vessel profiles. For biplane images, the location and orientation of the vessel segment will be determined using self-calibration technique for the determination of 3D vascular structure. Instantaneous and average blood flow rates in the selected vessel segments will be determined from analysis of the spatial shift of the distance- density curves of the vessel segment, which will be identified over a number of image frames based on the hierarchical vascular tree structure. Basic vessel information on vessel sizes, contrasts, and locations of opacified vessels will be obtained by means of automated tracking of vessel images based on the double-square-box region-of-search method. The hierarchical vascular tree structure will be constructed based on branching relationships between vessels and on basic vessel information. The 3D vascular structure in biplane angiography will be constructed without prior knowledge of the relationship between the two views. The relative geometry of the biplane imaging system will be determined from biplane image coordinates of eight or more object points of unknown 3D locations.
Showing the most recent 10 out of 15 publications
