Tracking Carotid Artery Wall Motion Using an Unscented Kalman Filter and Data Fusion

Abstract

Analyzing the motion of the common carotid artery (CCA) wall yields effective indicators for atherosclerosis. In this work, we propose a state-space model and a tracking method for estimating the time-varying CCA wall radius from a B-mode ultrasound sequence of arbitrary length. We employ an unscented Kalman filter that fuses two sets of measurements produced by an optical flow algorithm and a CCA wall localization algorithm. This fusion-and-tracking approach ensures that feature drift, which tends to impair optical flow based methods, is compensated in a temporally consistent manner. Simulation results show that the proposed method outperforms a recently proposed optical flow based method.

Analyzing the motion of the common carotid artery (CCA) wall yields effective indicators for atherosclerosis. In this work, we propose a state-space model and a tracking method for estimating the time-varying CCA wall radius from a B-mode ultrasound sequence of arbitrary length. We employ an unscented Kalman filter that fuses two sets of measurements produced by an optical flow algorithm and a CCA wall localization algorithm. This fusion-and-tracking approach ensures that feature drift, which tends to impair optical flow based methods, is compensated in a temporally consistent manner. Simulation results show that the proposed method outperforms a recently proposed optical flow based method.