Closed-line integral optimization edge detection algorithm and its application in equilibrium radionuclide angiocardiography.

Abstract

Automatic evaluation of left ventricular (LV) function using equilibrium radionuclide angiocardiography requires an edge detection algorithm to correct and reproducibly delineate the left ventricle. Available algorithms, usually based on differentiation of a radial profile, generally suffer from low precision due to low signal-to-noise ratios and overlapping structures, for example, the left atrium. An edge detection algorithm was developed based on the assumption that the LV border can be defined as the maximum, normalized, closed-line integral of a closed curve in a vector field derived by image differentiation. It is further assumed that the closed curve can be described by a Fourier expansion with a limited number of harmonics. Regions of interest (ROIs) generated by this algorithm were compared with ROIs generated by an algorithm based on a combination of thresholding and second-order derivatives. This algorithm delineates the left ventricle and gives results more closely related to ROIs generated manually than the algorithm combining thresholding and the second-order derivative. Our algorithm can also handle the problem of overlapping structures, as demonstrated in phantom simulations. The concept of a maximum, normalized closed-line integral will improve the delineation of the LV in an equilibrium radionuclide angiocardiography study. The problem of overlapping structures is overcome by this algorithm because it takes into consideration global edge information.