Sparse convolutional plane-wave compounding for ultrasound imaging

Ultrafast ultrasound imaging enables imaging at kilohertz frame rates, at the cost of a degraded image quality in terms of contrast and resolution. To reach an image quality comparable to focused imaging, the standard technique is coherent compounding, which requires a high number of transmissions and reduces the effective frame rate. In this paper, we introduce a COnvolutional Compounding Algorithm (COCA), a non-linear compounding method that aims at improving the quality of plane-wave imaging by virtually creating new angles. It enables to perform high-quality imaging with a significantly reduced number of emissions. Tests carried out on simulations, in vivo and in vitro experiments, using the Plane-wave Imaging Challenge in Medical UltraSound (PICMUS) dataset, show a significant improvement in terms of contrast and resolution compared to coherent compounding, increasing lateral resolution by 15 to 30 % and contrast by 6.8 dB in average, reaching an overall quality almost comparable to a coherent compounding of 75 angles with only 7 angles.