Options
2021
Conference Paper
Titel
Ultrafast adaptive coherence beamforming using a fast and simple quad phase estimator
Abstract
Ultrafast ultrasound imaging allows total focusing on every pixel, but it uses a non-focusing transmission of the ultrasonic wave into the tissue. Based on plane wave or diverging wave imaging, additional clutter reduces the imaging quality especially in echo-free regions like cysts. Clutter reducing techniques like adaptive beamforming processing compute weighting factors based on measured channel data used for each sample reconstruction. This increases signal and image quality again but adds computational complexity and may result in techniques that are not real-time capable to be implemented in an ultrasound scanner. A fast and accurate phase estimator for real-time applications is needed. Extending the concept of sign coherence factor (SCF) computations using bi-phase estimations, we propose a fast quad phase coherence (QPC) estimator for more accurate phase detection to reduce incoherent clutter. Instead of only using the sign coherence of each of the single channel data, the QPC estimator also includes the monotony of the signal to differentiate into the four classes "positive sign & increasing monotony", "positive sign & decreasing monotony", "negative sign & decreasing monotony" and "negative sign & increasing monotony". The dominant class of these four is selected and the adaptive weighting is computed as ratio of the absolute frequency of this class to the sum of all other class occurrences. The calculation of the monotony introduces only a small computational overhead compared to other adaptive beamforming techniques. The performance of QPC usage is evaluated using our ultrafast ultrasound research systems at phantom measurements (using ""DiPhAS"" system with 128 parallel channels) and in-vivo bladder measurements (using "MoUsE" system with 32 parallel channels, 3 MHz phased array transducer). The beamforming reconstruction is computed on the performance-limited integrated PC hardware as OpenCL kernel in real time. Imaging performance is increased compared to sign coherence beamforming (SNR by >3 dB, contrast by 4,4 dB, spatial resolution PSF FWHM by 19 % ) while computation time still allows real-time usage.