Ultrasound compressed sensing: Performance study of reconstruction on different ultrasound imaging data

In this work, we investigate performance of reconstruction on different ultrasound (US) post-beamformed data when performing compressed sensing (CS) with fewer measurements. Three types of post-beamformed data - radio frequency (RF), envelope, and log-converted data are used. Each data type has its own signal distribution, which, given a typical sparse representation, affects sparsifying efficiency during CS reconstruction and hence further determines restored image quality. Field II simulation of a speckle generating phantoms with a hyperechoic cyst is performed. The re-sampled image data used for evaluation of reconstruction performance has 75% data reduction rate from the original image. Here reconstruction is performed using convex optimization algorithm along the lateral direction and sparse transformation with CDF 5/3 wavelet is typically chosen. The results show that for feature recognition, reconstructions on post-beamformed RF and envelope data have close performance, and are better than that on log-converted data. In terms of mean-square-error, reconstruction on RF data performs best.