Adaptive Threshold for Eigenspace-Based Minimum Variance Beamformer for Dark Region Artifacts Elimination

Quality degradations caused by clutter often occur in medical ultrasound images when using nonadaptive delay-and-sum (DAS) beamforming method. A wide variety of beamforming methods have been studied to deal with this issue. Minimum variance (MV) beamformer represents a widely known method designed to minimize interference and noise power while retaining the desired signals in a distortion-free manner. However, it may be restricted in its clinical usage due to the poor contrast performance. Eigenspace-based MV (EMV) was formerly proposed to enhance the contrast of MV. However, this approach poses the problem that it usually introduces dark region artifacts alongside the hyperechoic scatterers when obtaining a high contrast, due to the fixed eigenvalue threshold in the subspace construction. First, an adaptive threshold is proposed to allow for a good tradeoff between dark region artifacts elimination and contrast enhancement. To go further, an adaptive spatially smoothed EMV (AEMV) is developed. In AEMV, adaptive selection of subarray length and adaptive threshold are intrinsically combined to separate and reduce OFF-axis energy effectively. Simulations, experiments, and
in vivo
studies were made to evaluate the performance of the proposed methods. Quantified results show that, compared with conventional EMV beamformer, AEMV yields the largest improvements in contrast ratio (CR), generalized contrast-to-noise ratio (GCNR), and speckle signal-to-noise ratio (sSNR) of 42.27%/90.88%, 18.07%/29.66%, and 64%/58.93%, respectively, in simulations and experiments. This demonstrates that AEMV is effective in improving image contrast while retaining the speckle pattern. The
in vivo
study also appears a potential for AEMV to detect lesions.