2019/07/24 by Shujaat Khan, Khan, Shujaat, Jaeyoung Huh +3
Engineering · Medicine · #Computer Vision and Pattern Recognition (cs.CV) #FOS: Computer and information sciences #FOS: Electrical engineering #Flow Measurement and Analysis #Image and Video Processing (eess.IV) #Machine Learning (cs.LG) #Photoacoustic and Ultrasonic Imaging #Ultrasound Imaging and Elastography #electronic engineering #information engineering
paper · pdf · doi:10.48550/arxiv.1907.10257
openalex publication_date 2019/07/24 · openalex created_date 2025/10/10 · openalex updated_date 2026/07/28
In ultrasound (US) imaging, various types of adaptive beamforming techniques\nhave been investigated to improve the resolution and contrast-to-noise ratio of\nthe delay and sum (DAS) beamformers. Unfortunately, the performance of these\nadaptive beamforming approaches degrade when the underlying model is not\nsufficiently accurate and the number of channels decreases. To address this\nproblem, here we propose a deep learning-based beamformer to generate\nsignificantly improved images over widely varying measurement conditions and\nchannel subsampling patterns. In particular, our deep neural network is\ndesigned to directly process full or sub-sampled radio-frequency (RF) data\nacquired at various subsampling rates and detector configurations so that it\ncan generate high quality ultrasound images using a single beamformer. The\norigin of such input-dependent adaptivity is also theoretically analyzed.\nExperimental results using B-mode focused ultrasound confirm the efficacy of\nthe proposed methods.\n