High-Quality Plane-Wave Image Reconstruction using the U2-Net Deep Learning Model

Plane-wave imaging (PWI) stands as the most prevalent ultrafast ultrasound imaging technique. Coherent plane-wave compounding (CPWC) merges PWI outcomes from steered plane waves at various angles to bolster image quality. Recently, adaptive beamformers have been studied to notably enhance the image quality of CPWC without compromising frame rate. However, adaptive beamformers often entail high computational complexity. In this study, we leverage the U²-Net deep learning model to discern the beamforming weights of the random matrix theory-based Wiener postfilter combined with eigenspace and spatial coherence-based minimum variance (RMTW-ESCMV) beamformer, expediting high-quality ultrasound image reconstruction. The proposed learnable RMTW-ESCMV (LRMTW-ESCMV) beamformer aims to swiftly reconstruct high-quality adaptive beamforming images from single-angle PWI images. Compared to RMTW-ESCMV, the LRMTW-ESCMV beam-former has excellent computational efficiency while achieving comparable image quality.