Ultrasound image quality optimization with adaptive global sound speed correction

Imaging with correct sound speed is essential to ensure beamforming accuracy and thus final image quality in diagnostic ultrasound. In most clinical systems, however, a predetermined constant sound speed is used, so the mismatch between the presumed value and the actual sound speed distribution is likely to degrade both spatial and contrast resolution. In this paper, a global sound speed estimation method is proposed to optimize beamforming quality of a phased array system. Previous approaches of global sound speed estimation are based on the minimum average phase covariance method or the speckle auto-covariance function method to estimate sound speed in an iterative fashion. It was found that large errors may still be present. The proposed method here utilizes a sub-aperture imaging technique to form two images, and the mean error between the two images is minimized by choosing an optimal sound speed value. This can be implemented with a parallel architecture to speed up the computations. The optimal speed from the proposed method provides better CNR on all our experimental data, including those from standard tissue mimicking phantoms, aberration phantoms as well as clinical breast data. With our proposed method, the sound speed estimation error is about ±0.7% (±10m/s), which is better than ±2.1%(±30m/s) with the auto-covariance function method and ±2.7%(±40m/s) with the minimum average phase covariance method. Thus, the proposed method has better performance in automatic global sound speed estimation than the conventional methods. In addition, it can be efficiently implemented in software. On a GPU platform with CUDA, the computations can be performed within 1 sec.