Motion Estimation in Echocardiography Using Sparse Representation and Dictionary Learning

Abstract

OATAO is an open access repository that collects the work of Toulouse researchers and makes it freely available over the web where possible. Abstract-This paper introduces a new method for cardiac motion estimation in 2-D ultrasound images. The motion estimation problem is formulated as an energy minimization, whose data fidelity term is built using the assumption that the images are corrupted by multiplicative Rayleigh noise. In addition to a classical spatial smoothness constraint, the proposed method exploits the sparse properties of the cardiac motion to regularize the solution via an appropriate dictionary learning step. The proposed method is evaluated on one data set with available ground-truth, including four sequences of highly realistic simulations. The approach is also validated on both healthy and pathological sequences of in vivo data. We evaluate the method in terms of motion estimation accuracy and strain errors and compare the performance with state-of-the-art algorithms. The results show that the proposed method gives competitive results for the considered data. Furthermore, the in vivo strain analysis demonstrates that meaningful clinical interpretation can be obtained from the estimated motion vectors.

Robust Optical Flow Estimation in Cardiac Ultrasound Images Using a Sparse Representation

“…The considered energy is defined as the sum of a data fidelity term denoted as E Q and regularizations denoted as E S and E W . The first regularization E S ensures a smooth spatial variation of the motion field, while the second one E W exploits the patchwise sparse properties of the motion vectors in U, when decomposed on a learnt dictionary D [21], [22]. The motion field is obtained through the minimization of the resulting cost function…”

“…where α is the sparse coefficient vector, λ d and λ s are two regularization parameters that control the influence of the two regularizations. Prior to the motion estimation, the motion dictionaries are learnt offline from a set of training cardiac motion fields as in [21]. In a second step, the motion of each pair of test images is estimated using the minimization problem (4).…”

“…The proposed motion estimation method is based on a recent work that has shown the benefits of a sparsity-based prior for cardiac motion estimation [21]. In addition to exploiting the sparse regularization promoted in [21], this paper investigates a new motion estimation method mitigating the effect of outliers. The problem is formulated within a general OF-based energy minimization framework.…”

“…The problem is formulated within a general OF-based energy minimization framework. As in [21], two regularization terms are used to enforce spatial smoothness and sparsity of motion in a learnt cardiac motion dictionary. Moreover, robustness is introduced using weighting functions derived from M-estimators.…”

“…In Section IV, synthetic experiments are used to investigate the interest of robustness for the data fidelity and regularization terms. The proposed method is then compared to the previous non-robust cardiac motion estimation method of [21] and to the robust OF method of [26], using realistic simulations of cardiac images. The feasibility of the method for real data is also demonstrated using in vivo images.…”