Longitudinal motion of the carotid artery wall and speckle decorrelation issue: accurate estimation using a Kalman-based speckle tracking method

This work presents an original 2D motion estimation framework, applied to the carotid artery in ultrasound B mode sequences. Our method is based on a specific block matching approach, in which the reference block is updated in each frame by a Kalman filtering scheme. Applied in vivo to a cohort of 28 subjects, our method showed a high robustness and accuracy, with a mean absolute error of 47 +/- 39 mu m and 18 +/- 14 mu m for the trajectory estimation, and 55 +/- 55 mu m and 26 +/- 22 mu m for the amplitude estimation, in the longitudinal and radial directions, respectively. This corresponded to the magnitude of the inter-and intra-observers variability, and was significantly inferior in comparison with two other methods, i.e. a classical block matching implementation, and the Velocity Vector Imaging commercial software. Our method was also fully reproducible as the resulting variability was null. The proposed framework may constitute a reliable method to assess the arterial stiffness.