Higher Order Singular Value Decomposition Filter for Contrast Echocardiography

Abstract

Assessing the coronary circulation with contrast-enhanced echocardiography has high clinical relevance. However, it is not being routinely performed in clinical practice because the current clinical tools generally could not provide adequate image quality. The contrast agent's visibility in the myocardium is generally poor, impaired by motion and non-linear propagation artifacts. The established multi-pulse contrast schemes (MPCS) and the more experimental singular value decomposition (SVD) filter also fall short to solve these issues. Here, we propose a scheme to process AM/AMPI echoes with higher-order singular value decomposition (HOSVD) instead of conventionally summing the complementary pulses. The echoes from the complementary pulses form a separate dimension in the HOSVD algorithm. Then, removing the ranks in that dimension with dominant coherent signals coming from tissue scattering would provide the contrast detection. We performed both in vitro and in vivo experiments to assess the performance of our proposed method in comparison with the current standard methods. A flow phantom study shows that HOSVD on AM pulsing exceeds the contrast-to-background ratio (CBR) of conventional AM and an SVD filter by 10dB and 14dB, respectively. In vivo porcine heart results also demonstrate that, compared to AM, HOSVD improves CBR in open-chest acquisition (up to 19dB) and contrast ratio in closed-chest acquisition (3dB).