Medical ultrasound imaging using neural networks

A neural network can be trained to remove aberrations in the echo arrival time (delay noise) prior to beamforming in a medical ultrasound imaging system. Such aberrations arise from the inhomogeneity in the velocity of sound in the intervening tissues and can severely degrade the image quality. Simulations for a 64-element, 3.5 MHz array coupled with a 64-input/output neural network demonstrate the improvements.