|Ph.D Student||Chernyakova Tatyana|
|Subject||Low-Rate Generalized Beamforming for Medical Ultrasound|
|Department||Department of Electrical Engineering||Supervisor||Professor Yonina Eldar|
Ultrasound imaging includes brightness, motion, Doppler, elastography and contrast enhanced modalities. Imaging is performed with an array of transducer elements allowing for spatial selectivity of signal transmission and reception. In delay-and-sum (DAS) beamforming, used in all commercial systems today, the delayed signals are averaged to yield the signal of interest originating in a certain point. When implemented digitally, DAS requires severe oversampling of the detected signals. As the number of transducer elements can be as large as several thousand, huge amounts of data need to be transmitted from the system front-end and digitally processed in real time. Data loads limit implementation of novel imaging techniques developed to improve image quality and accelerate acquisition.
In this work we study main limitations of DAS processing and propose alternatives for more efficient implementation, integration of advanced imaging paradigms and improved image quality.
In the first part of this work we show that oversampling can be avoided by translating the beamforming to the frequency domain. Furthermore, sampling and processing can be performed at rates lower than Nyquist, when the structure of the beamformed signal is exploited through a compressed sensing framework. We implement the proposed method, frequency domain beamforming (FDBF), for 2D and 3D focused imaging, as well as for coherent plane-wave compounding allowing for ultrafast acquisition. We next exploit the frequency domain processing for efficient implementation of pulse compression required for imaging with coded transmissions.
In addition to limitations posed by existing implementation, DAS beamforming has intrinsic effect on image quality. Due to the finite aperture size, the off-axis echoes are not entirely suppressed and are expressed as the main lobe width and side-lobe level of the DAS beampattern.
A statistical interpretation of beamforming allowing to overcome the limitations of DAS processing is presented in the second part of this work.
We propose a scheme iterating estimation of distribution parameters and computation of maximum-a-posteriori (MAP) estimator of the signal of interest, leading to an iterative MAP beamformer. It yields significantly improved contrast compared to DAS without severely increasing computational complexity or need for fine-tuning of parameters.