Ultrasound Images Using a Motion Estimation Algorithm

This study estimates flow patterns of contrast agents from successive ultrasound image sequences by using an anisotropic diffusion-based optical flow algorithm. Before flow fields were recovered, the test sequences were reconstructed using relative composition of structural and textural parts from the original image. To improve estimation performance, an anisotropic diffusion filtering model was embedded into a spline-based slightly nonconvex total variation-L1 minimization algorithm. In addition, an incremental coarse-to-fine warping framework was employed with a linear minimization scheme to account for a large displacement.

After each warping iteration, the implementation used intermediate bilateral filtering to prevent oversmoothing across motion boundaries. The performance of the proposed algorithm was tested using three different sequences obtained from two simulated datasets and phantom ultrasound sequences. The results indicate the robust performance of the proposed method under different noise environments. The results of the phantom study also demonstrate reliable performance according to different injection conditions of contrast agents. These experimental results suggest the potential clinical applicability of the proposed algorithm to ultrasonographic diagnosis based on contrast agents.