Phase-Based Heart Motion Estimation in Multimodality Cardiac Imaging

International audienceThis chapter presents two methods for motion estimation. After an introduction to the concepts of hypercomplex signals and of spatial phase, the two methods proposed are described and applied to cardiac image sequences. The first motion estimation method uses the spatial phases of MR images calculated with two 2D SOASs. The second method is based on a formulation of the optical flow that makes use of the properties of monogenic signals, on a local affine modeling of motion and a multiscale analysis that enables the local selection of the most appropriate motion estimation. The chapter proposes a comparison between Sühling's algorithm and the algorithm based on the extracted phase of the monogenic signal exploiting either a motion model by translation or an affine motion model. The result proves to reinforce the idea that the hypothesis of phase conservation during the cardiac cycle is very well suited for ultrasound imagin