Investigation on improving performance of adaptive beamformer by statistical analysis of ultrasonic echoes

Abstract

Minimum variance (MV) beamformers have been introduced in medical ultrasound imaging to improve image quality. In most cases, the MV beamformers have been investigated in terms of resolution improvement. However, the contrast-to-noise ratio (CNR) is also a clinically important metrics and gathers attention recently. In this study, we examined the diagonal loading parameter σ in MV beamforming and determined its appropriate value by evaluating image quality evaluation metrics including CNR. In order to further improve the image quality, a method for determining the value of σ based on the difference in statistical properties of received ultrasonic echo signals was also investigated. The phantom experimental results showed that the proposed method achieved a better CNR than the conventional MV beamformer while keeping resolution significantly better than that in delay-and-sum (DAS) beamforming.

Effects from imaging parameters of sub-aperture averaging in minimum variance adaptive beamforming using outputs of delay-and-sum beamformer

“…We proposed MV adaptive beamforming based on the statistical analysis of ultrasonic echoes. 15) There are some researches using strategies except for the MV method to improve the image qualities of B-mode images. [16][17][18][19] Li et al 16) developed an adaptive weighting strategy using the CF 14) to reduce an error in the focusing.…”

Beamspace-domain learning of minimum variance beamformer with fully convolutional network

“…We also tried to improve the performance of the MV beamformer with subaperture beamforming, 6) apodized or cross-covariance matrix, 7,8) weighting by the phase coherence factors 9) and introducing speckle statistics. 10) MV beamforming provides higher resolution than DAS beamforming, but the computational cost of matrix inversion is high and real-time processing is difficult.…”

“…(4). 10) However, no study on s in MV beamforming with the cross-covariance matrix has been reported. Therefore, in this study, we examined the optimal s for the cross-covariance matrix.…”