High Contrast Minimum Variance Beamforming Combined with Convolutional Neural Network

This work combines convolutional neural network and traditional minimum variance (MV) beamforming to produce ultrasound images with higher image contrast. The received ultrasound channel data are in time domain, and the designed deep learning neural network is based on convolutional neural network. The network model is responsible for suppressing off-axis scattering signals, and the apodization weights of MV beamforming guarantee the image resolution performance. Experimental results demonstrated that the proposed method significantly improved the contrast performance of MV beamforming while reserving high lateral resolution.