Trained compounding operator to improve ultrasound imaging frame rate

Plane and diverging wave imaging uses unfocused transmits to ensonify a larger field of view with transmitted pulses thus achieving higher frame rate compared to focused transmits. The loss of intensity in the transmitted pulse and penetration at depth compared to focused transmits is mitigated by coherently compounding several frames together at tilted transmit angles. However, this limits the available increase in frame rate. In this work, we investigate non-linear compounding operators using neural nets to achieve improved image quality with fewer angles, thereby reducing the impact of this tradeoff. The proposed method improves contrast of images using a two-stage compounding process operating on acquired transmit data in a sliding queue. The contrast is increases by approximately 6.9 dB in silico at the first stage and by a further 7.5 dB at the second stage. In vivo the contrast increases by 4.2 dB at the first stage and by 7.1 dB at the second stage.