Eigen-based signal processing methods for ultrasound color flow imaging

Purpose. This thesis presents a study on the design of new eigen-based signal processing methods for use in color flow imaging. Specifically, the proposed methods are designed to address three main problems in color flow signal processing: the lack of abundant Doppler data samples, the possible presence of wideband Doppler clutter, and the potential flow signal distortions that arise during clutter suppression.; Theoretical contributions. A clutter filter (the Hankel-SVD filter) and a flow estimator (the Matrix Pencil method) were respectively developed by exploiting the eigen-space principles related to two matrix forms known as the Hankel matrix and the matrix pencil. Both techniques were then combined to form a new color flow data processor that performs flow detection and flow estimation in parallel. All of these methods are adaptive to the Doppler signal contents through an SVD of the Hankel matrix created from a Doppler data vector. Moreover, they are intended to work with each Doppler ensemble separately (i.e. they do not require Doppler ensembles from multiple sample volumes).; Simulation assessment. A Doppler signal simulation model was developed to generate Doppler signals consisting of non-stationary (phase-modulated) tissue clutter and spectral-broadened (amplitude-modulated) flow echoes. The synthesized datasets were used to analyze the Hankel-SVD filter's flow detection performance and the Matrix Pencil's velocity estimation performance. A comparison of the Hankel-SVD filter with an existing type of adaptive filter (clutter-downmixing filter) showed that this new filter is more capable of discriminating flow echoes from Doppler clutter. Also, it was found that the Matrix Pencil can provide less biased flow estimates as compared to other frequency-based flow estimators like the lag-one autocorrelator.; Experimental assessment. Color flow datasets were obtained respectively from a steady-flow phantom and from the carotid arteries of a youth subject. These datasets were used to analyze the performance of the proposed single-module processor. For the flow phantom study, the single-module processor showed that it can reconstruct velocity maps similar to the theoretical flow profile. As well, for the in vivo studies, it gave a better flow detection performance than conventional two-module processors especially when there is substantial tissue motion.