Fundamental studies on boundary estimation of low SNR arteriogram by 'minimum difference of density method'

Boundaries have usually been detected by using the so-called spatial differentiation method. This method is accomplished by measuring the magnitude of density transitions at each regions. So, in case of the low SNR image such as angiogram by IVDSA method, false detection of boundaries ensued because of emphasized noise. In this study, we were able to trace the opacified blood vessel boundaries from such images by using stochastic technique and looking for variance of the integrated density along the vessel. Since one of the most significant properties of the blood vessels lies in their continuity, prior knowledges of the states (direction, diameter, etc) of the adjoing structure are useful to detect true boundaries of the low SNR opacified vessels. The performance of the proposed method (Minimum Difference of Density method or MDD method) is demonstrated by applying the new algorithm to both a simulated vessel and a human coronary arteriogram obtained by the Judkins' method. This method can be expected be applied in low SNR coronary arteriogram by using IVDSA method. (author)