A sequential Bayesian based method for tracking and strain palpography estimation of arteries in intravascular ultrasound images

This paper investigates the task of tracking and strain estimation of arteries in intravascular ultrasound images. A tracking method is proposed to extract the inner and the outer contours of the vessel wall (lumen/intima-media and intima-media/adventitia interfaces, respectively), and the deformations along them. This estimation is carried out by a non parametric sequential Bayesian method. The Bayesian modeling holds three ingredients: the prior, which is given by a manually defined segmentation of the contours on the first image; the transition, which is assumed to follow a Markovian random walk; and the likelihood, which is a distance between patches distributed along the contours. The underlying Bayesian posterior distribution is approximated using a sequential Monte Carlo approach. Experiments on three PVA-C phantoms present direct readings of the deformations along the lumen/intima-media contour.