IEEE Trans Ultrason Ferroelectr Freq Control. 2020 Aug;67(8):1497-1512. doi: 10.1109/TUFFC.2020.2975483. Epub 2020 Feb 20.
Blind source separation (BSS) refers to a number of signal processing techniques that decompose a signal into several "source" signals. In recent years, BSS is increasingly employed for the suppression of clutter and noise in ultrasonic imaging. In particular, its ability to separate sources based on measures of independence rather than their temporal or spatial frequency content makes BSS a powerful filtering tool for data in which the desired and undesired signals overlap in the spectral domain. The purpose of this work was to review the existing BSS methods and their potential in ultrasound imaging. Furthermore, we tested and compared the effectiveness of these techniques in the field of contrast-ultrasound super-resolution, contrast quantification, and speckle tracking. For all applications, this was done in silico, in vitro, and in vivo. We found that the critical step in BSS filtering is the identification of components containing the desired signal and highlighted the value of a priori domain knowledge to define effective criteria for signal component selection.
盲源分离（BSS）是指将信号分解成几个“源”信号的一系列信号处理技术。近年来，BSS 越来越多地用于抑制超声成像中的杂波和噪声。特别是，BSS 能够基于独立性度量而不是其时间或空间频率内容来分离源，这使其成为数据的强大滤波工具，其中所需和不需要的信号在频谱域中重叠。本工作的目的是回顾现有的 BSS 方法及其在超声成像中的潜力。此外，我们还测试和比较了这些技术在对比超声超分辨率、对比量化和斑点跟踪领域的有效性。对于所有应用，这都是在计算机模拟、体外和体内进行的。我们发现，BSS 滤波的关键步骤是识别包含所需信号的分量，并强调了先验域知识的价值，以定义信号分量选择的有效标准。
