Deep Learning-based Ultrasound Computed Tomography for Cortical Bone Imaging

Chenchen Zhou,Tao Jiang,Kailiang Xu,Dean Ta

DOI: https://doi.org/10.1109/IUS51837.2023.10307403

2023-01-01

Abstract:Ultrasound computed tomography (UCT) utilizes parameter inversion for tissue image reconstruction and has emerged as a novel technology for cortical bone imaging in recent years. However, traditional UCT algorithm, such as FWI, suffers from computational burden and inversion instability. Deep learning (DL) has been applied in seismic FWI for fast and accurate reconstruction of subsurface structures. In the present study, a data-driven DL-based UCT (DL-UCT) method is proposed for cortical bone imaging. Simulation demonstrates that the DL-UCT can reconstruct a bone image within 2 seconds and remain effective even in the absence of low-frequency information. The efficiency and stability of this method may surpass that of FWI.