Two New Methods for Finding Endocardial and Epicardial Boundaries in Echocardiographic Images Using Wavelet Analysis

In this paper two new algorithms for left ventricular (LV) endocardial and epicardial boundary detection from echocardiographic images are presented. The proposed algorithms are Center-Based approach in which the LV center point (LVCP) is estimated using a fuzzy-Based technique and then epicardial and endocardial edge points are searched on radial lines emanating from LVCP. A basic assumption in the first algorithm is that the boundary points of left ventricle are those with sharpest intensity changes, which can be detected as the global maximum wavelet transform modulus. In the second algorithm directional gradient with LVCP references are computed in all directions and all scales using wavelet transform. The images of gradient at different scales are then combined by logical operation to retain only possible boundary points. Mathematical morphology operations are applied to fill the dropouts in the resulting image. Radial search method is then used to find the boundary points. After de-noising by wavelet transform, cubic Bspline approximation method is used to achieve smooth and closed endocardial and epicardial boundaries. Sample experimental results are shown for endocardial and epicardial border identification in 2D short axis (SA) echocardiograms.