A fast and high frame rate adaptive beamforming using DCT‐based RF‐line recovery in line‐by‐line ultrasound imaging

Abstract

Ultrasound imaging is an important modality used in medical imaging. One of the significant stages in the ultrasound imaging is the beamforming process. This article proposes a new technique for reducing the overall computational time of adaptive linear ultrasound imaging. The method uses the discrete cosine transform‐based reconstruction for missing data imputation. The novelty of the paper is that we do not need to beam‐form the total scan lines, so the time of image construction can be saved significantly. In other words, a fraction of the total scan lines is selected for beamforming and the others are assumed to have values as Not‐a‐Number (NaN). The proposed reconstruction technique tries to assign appropriate values to the NaN ones. We applied the proposed method to the simulated and experimental radio frequency (RF) datasets for resolution and contrast evaluation. Results showed that the proposed technique is near to the minimum variance (MV) method in terms of resolution and contrast, and has less computational time for image formation compared to the MV. As some quantitative examples in some experiments we have formed only 50% and 33% of the total lines and reconstructed the rest, then we have been able to increase the frame rate twice and three times, respectively, which can be very useful in many applications, especially in echocardiography imaging. In addition, since the execution time of the reconstruction algorithm is not very significant, we were also able to increase the speed by two and three times while achieving an error of less than 10% compared to the case of using all image lines.

Synthetic Aperture Ultrasound Imaging through Adaptive Integrated Transmitting-Receiving Beamformer

“…20 It was also proposed to apply MV beamforming only on a fraction of total scan lines to reduce the computational time. 21 Kim et al 22 proposed to approximate the MV weights as a linear combination of a few principal components estimated through offline calculations. Park et al 23 represented a low complexity implementation of MV beamformer through fast QR decomposition, but their method is not applicable when the temporal averaging is used.…”