Ultra-fast ultrasound imaging relies on coherent Plane Wave (PW) compounding to obtain sufficient spatial resolution, and contrast. However, the process of coherent PW compounding incurs a loss in temporal resolution. We propose a Deep Learning (DL) network that achieves high resolution PW compounding using a reduced number of PW transmits. We embed a model based signal processing algorithm in the design of the network, which leads to better performance through the exploitation of the prior information that is now available to the network. Our proposed method outperforms two benchmark networks, yielding approximately an 8.2% improvement in PSNR, over the next best network. Aiming for an additional boost in resolution, we moreover train towards images acquired using higher transmit frequencies.
|Title of host publication||IUS 2020 - International Ultrasonics Symposium, Proceedings|
|Publication status||Published - 2020|
- Machine Learning
- Plane Wave Compounding
- Signal Processing