Estimating contrast agent motion from ultrasound images using an anisotropic diffusion-based optical flow technique

The purpose of this study was to suggest a novel optical flow method to estimate the motion patterns of contrast agents from an ultrasound image. We first recomposed the original image with relative structural and textural parts. We embedded an anisotropic diffusion model into a slightly non-convex total variation-L1 approximation scheme to provide more reliable estimates. An intermediate bilateral filter was adopted after each computation step to prevent over-smoothing effects. Ultrasound data were acquired during continuous injection of contrast agent into a tissue mimicking phantom. The results showed that our method provided robust performance for estimating contrast agent flow patterns.