Frequency-based signal processing for ultrasound color flow imaging

Abstract

In ultrasound color flow imaging, the computation of flow estimates is well-recognized as a challenging problem from a signal processing perspective. The flow visualization performance of this imaging tool is often affected by error sources such as the lack of abundant signal samples available for processing, the presence of wideband clutter in the acquired signals, and the flow signal distortions that may arise during clutter suppression. In this article, we review existing frequency-based signal processing approaches reported in the ultrasound literature and evaluate their theoretical advantages as well as limitations. In particular, four major classes of clutter filter designs are considered: FIR/IIR filtering, polynomial regression, clutter-downmixing, and eigen-regression. Also, three types of frequency estimators are discussed: lag-one autocorrelation, autoregressive modeling, and MUSIC. In examining these approaches, it was concluded that eigen-based methods like the eigen-regression filter and the MUSIC estimator can better adapt to the Doppler signal characteristics, and thus they seem to have more potential for obtaining flow estimates that are less affected by the signal processing error sources.