Deep Learning in RF Sub-sampled B-mode Ultrasound Imaging
share
In portable, three dimensional, and ultra-fast ultrasound (US) imaging systems, there is an increasing need to reconstruct high quality images from a limited number of RF data from receiver (Rx) or scan-line (SC) sub-sampling. However, due to the severe side lobe artifacts from RF sub-sampling, the standard beam-former often produces blurry images with less contrast that are not suitable for diagnostic purpose. To address this problem, some researchers have studied compressed sensing (CS) to exploit the sparsity of the image or RF data in some domains. However, the existing CS approaches require either hardware changes or computationally expensive algorithms. To overcome these limitations, here we propose a novel deep learning approach that directly interpolates the missing RF data by utilizing redundancy in the Rx-SC plane. In particular, the network design principle derives from a novel interpretation of the deep neural network as a cascaded convolution framelets that learns the data-driven bases for Hankel matrix decomposition. Our extensive experimental results from sub-sampled RF data from a real US system confirmed that the proposed method can effectively reduce the data rate without sacrificing the image quality.
READ FULL TEXT