SNR-dependent coherence weighting for minimum variance beamforming

Coherence factor (CF) is used to suppress sidelobes and to enhance contrast in ultrasound imaging. Recent research also introduces minimum-variance-distortionless-response (MVDR) based CF (CF_mvdr) to achieve good resolution for high frame rate application. However, increasing frame rate by unfocused beams reduces signal-to-noise-ratio (SNR). CF-based weighting is susceptible to noise, which underestimates the acoustic strength and lowers the contrast supposed to provide. Wiener postfilter, with worse performance on noise reduction but better robustness, can be combined with CF by scaling the noise power. Hence, CF weighting can be modified so as to avoid the noise effect with the knowledge of local SNR. The proposed method is specifically applied to the high frame rate imaging that has synthetic transmit aperture (STA) and MVDR beamformer. Before finding the MVDR solution, the channel data are used to estimate SNR, which then determines the noise power weighting by a sigmoid function and modifies the CF value. Specifically, at mainlobe with relatively high SNR, the CF or the CF_mvdr is adjusted by lowering its noise weighting. For the region requiring higher resolution, MVDR beamformer can retain its performance. Thus, the contrast can be restored when the noise level is high. Simulations were performed by a 128-element linear array with central frequency 4.37 MHz. The pulse-echo data were generated through STA method with eight subgroups. When the channel has SNR -10 dB, the CR and the CNR by original CF/CF_mvdr were 39.01/34.71 dB and 2.00/1.74 respectively. After modified by SNR, the CR and the CNR were improved to 47.45/47.82 dB and 2.26/2.14 in scaled CF/CF_mvdr cases. Simulations for point pairs was also included to confirm the ability to maintain resolution. In summary, a robust method designed for low SNR cases is proposed.