Deep Learning Beamforming for Sub-Sampled Ultrasound Data

In medical imaging tasks, such as cardiac imaging, ultrasound acquisition time is crucial, however traditional high-quality beamforming techniques are computationally expensive and their performance is hindered by sub-sampled data. To this end, we propose DeepFormer, a method to reconstruct high quality ultrasound images in real-time on sub-sampled raw data by performing an end-to-end deep learning-based reconstruction. Results on an in vivo dataset of 19 participants show that DeepFormer offers promising advantages over traditional processing of sub-sampled raw-ultrasound data and produces reconstructions that are both qualitatively and visually equivalent to fully-sampled DeepFormed images.