Abstract

One of the major advantages of ultrasound is its ability to image at very high frame rates, which can be exploited to track tissue motion. In this thesis, we focus on two important applications of motion estimation, namely ultrasound elastography and clutter suppression. In both of these applications, the tracking problem poses several technical challenges and as such is an active field of research. In elastography, tracking motion while a tissue undergoes some deformation reveals the physiological condition of the tissue by mapping its mechanical properties. We process ultrasound Radio-Frequency (RF) frames acquired before and after tissue deformation to estimate tissue displacement and eventually tissue strain. We propose a novel ultrasound elastography method where unlike conventional techniques, three ultrasound RF frames are taken into account to devise a cost function consisting of data term, spatial regularization terms and temporal continuity prior. We find the strain map by taking the spatial derivative of frame to frame displacement field estimated by efficient optimization of the aforementioned cost function. Validation with simulation, phantom and in-vivo liver data shows that the proposed technique substantially outperforms the state-of-the-art ultrasound elastography algorithms in terms of conventional quality metrics such as Signal-to-Noise Ratio (SNR), Contrast-to-Noise Ratio (CNR) and Strain Ratio (SR).

In clutter suppression, enhancement of blood flow by suppressing the clutter (i.e. non-moving stationary tissue) components is vital for assessing vascular health. In this thesis, a novel technique for suppressing clutter in ultrasound Color Flow Imaging (CFI) has also been proposed. Since the state-of-the-art Singular Value Decomposition (SVD) based technique is highly dependent on the proper selection of the boundaries between different subspaces, it is prone to producing nonoptimal clutter suppressed power Doppler images. In addition, extensive manual intervention typically needed to find the correct subspace ranks makes SVD difficult to be implemented on clinical ultrasound machines. To overcome these limitations, we propose to look at the clutter suppression problem from the standpoint of separating the foreground from the background. Precisely, we adapt the fast Robust Matrix Completion Algorithm (fRMC) where the in-face extended Frank-Wolfe method has been taken into account to decompose the Casorati matrix into low rank clutter and sparse blood components without requiring any manual tuning. We validate the proposed algorithm with simulation, experimental flow phantom, in-vivo animal and human datasets to show that our technique confidently attains the optimal result without requiring any manual intervention.