2D local heart motion estimation using level sets and hierarchical B-splines

Diagnosing complex heart malformations require the analysis of local heart motion. This paper presents a method for measuring local motion in a sequence of 2D echographic heart images. The heart cavities are segmented using an implicit geometric active contour robust to speckle. A combined rigid and non-rigid image registration method is then used to estimate the motion field. The cavity segmentation is done with our original level set method founded on the coefficient of variation. The curves obtained from two successive images are matched during the registration process to estimate their displacement. The registration process consists of three phases. First, we estimate a linear transform using the ICP algorithm to remove the linear difference between the cavities. The second phase consists of calculating an initial deformation embedding the B-spline transform in the ICP algorithm. Finally, we perform the hierarchical B-spline refinement in regions with unsatisfactory deviations. The result is a dense free form deformation field that is segmented to show the local heart motion. Experimental results using real ultrasound images are shown.