Bayesian-based beamformer with wiener post-filter for adaptative processing of ultrasond image using coherent plane wave compounding

Abstract

Conventional Ultrasound (US) scanners still implement the non-adaptive Delay and Sum (DAS) technique due to easy implementation and low processing time, but it provides low resolution and image contrast. Adaptive techniques allow obtaining high resolution and high contrast at the cost of complex implementation and high processing time. In this work we suggest the implementation of a Bayesian Minimum Variance (MV-BY) beamformer combined with the Wiener post-filter (WPF) to form the MV-BY-WPF beamformer. The MV-BY-WPF results in a combination of the proposed Bayesian-based Minimum Variance (MV-BY) beamformer with the WPF for adaptive beamforming of US image using Coherent Plane Wave Compounding (CPWC). The CPWC imaging has low Signal-to-noise Ratio (SNR) due to the lack of focusing on the pulse emission. Different adaptive methods such Coherence Factor (CF) or WPF beamformer have been suggested to overcome such limitation, however they still present limitations. In this study, we have introduced the MV-BY beamformer which takes subtle advantages over the Minimum Variance (MV) beamformer by applying a Bayesian-based post-filter which improves the performance of the MV. The proposed MV-BY is combined with the WPF to form the MV-BY-WPF which outperforms the MV-WPF. The MV-BY-WPF provides better contrast while retaining the imaging brightness at comparable spatial resolution when compared to the beamformers such as the traditional MV-WPF. Using our proposed methods, improvements have been introduced in CPWC imaging with comparable computational complexity with the MV beamformer. The performance evaluation of the proposed techniques includes spatial resolution, contrast, and speckle statistics. Particular to our proposed methodology, we have found that approximately 5 emissions using available data for research in US imaging, we have reconstructed an image with quality comparable to that obtained when using a total amount of 75 emissions with DAS beamformer. The reconstructions using simulation, phantom, and in-vivo dataset reveal the effectiveness of our proposed beamformer in terms of array noise suppression abilities. For example, using human in-vivo data, when DAS is compared to the MV-BY/MV-BY-WPF values in percentage (%) of 9.34/17.65 and 16.18/31.79 for Contrast Ratio (CR) and Contrast-to-noise Ratio (CNR), were obtained. Additionally, the MV-BY-WPF better retains the speckle when compared to MV-WPF. This means that our proposed methodology can improve the imaging system, and is suggested for real applications.