M.Sc Thesis

M.Sc StudentRabinovich Adi
SubjectAdaptive Beamforming in High Frame Rate Medical
Ultrasound Imaging
DepartmentDepartment of Electrical and Computer Engineering
Supervisors PROFESSOR EMERITUS Arie Feuer
DR. Zvi Friedman


In recent years, a variety of methods that increase frame rate of ultrasound video sequence have been proposed. The common feature of them all is generating multiple receive lines from one or more transmit events, so that the number of receive lines is several times larger than the number of transmits. This is especially important for cardiac diagnostic imaging, where it is desirable to be able to examine the temporal behavior of fast phases in the cardiac cycle.

This work concentrates on two major approaches at increasing ultrasound imaging frame rate, Multi-Line Acquisition (MLA), and Multi-Line Transmission (MLT). In Multi-Line Acquisition a wide transmission beam spans several receive lines, which comes at the cost of a deterioration of the lateral resolution of the ultrasound image. An MLA scan also induces block-like artifacts in the image which require proper compensation, by a method such as Synthetic Transmit Beamforming (STB). This in turn either reduces the potential gain in frame rate, or deteriorates lateral resolution even more. In Multi-Line Transmission the transmission is simultaneously focused at several directions. This introduces image artifacts known as cross-talk. So, both methods seem to trade-off an increase in frame rate with image quality deterioration.

Adaptive beamforming, a data-dependent approach to beamforming, has been applied to medical ultrasound imaging, and received growing interest in recent years. It was demonstrated that adaptive beamforming can provide superior image quality in terms of spatial resolution, signal-to-noise, and contrast-to-noise ratios. The main goal of our study was to incorporate the adaptive beamforming technology, and specifically Minimum Variance (MV) beamforming, in either the MLA, or the MLT processes. We demonstrate that while not sufficient by itself in MLA scanning, when combined with STB compensation, it results indeed in major improvement of image quality. We also demonstrate similar results when adaptive beamforming is applied to MLT scanning. Since the application of MV beamforming is computationally intensive, it cannot be applied in current systems. This requires the development of approximation methods we present here, including our original proposal for low-complexity approximation of the MV beamformer.

Simulation results, as well as experimental phantom and cardiac data, demonstrate the feasibility of improving lateral resolution for MLA scans, and reducing cross-talk artifacts for MLT scans, with adaptive beamforming. While, achieving superior spatial resolution, not only over other MLA and MLT methods with data-independent beamforming, but even over that of lower frame rate, Single-Line Acquisition (SLA) data-independent beamformers.