Increasing Frame Rate of Focused Ultrasound Imaging Based on Tensor Completion

This study proposes a method for improving the ultrasound imaging frame rate during the pre-processing step. Tensor completion (TC) is a powerful technique for retrieving missing information from a low-rank tensor, based on rank minimization and a minimal number of observations. Here, we apply this technique to traditional focused ultrasound imaging (TFUI). The concept behind this study is to beamform a percentage of scan lines that are randomly selected in each frame using a novel binary coding pattern. Furthermore, because the combination of known and unknown information yields an incomplete tensor, we suggest the TC technique for estimating the tensor (i.e., for the reconstruction of missing scan lines). We tested the proposed technique on a pre-beamformed radiofrequency (RF) ultrasound dataset of a human heart. The results demonstrate that we can achieve a 4x frame rate increase while the structural similarity (SSIM) between reconstructed and original tensors is greater than 0.98, which indicates that the information lost due to the application of TC is less than 2%.