Impact Factor:3.694

DOI number:10.1016/j.ultrasmedbio.2022.05.033

Journal:Ultrasound in medicine and biology

Abstract:Ultrasound sound-speed tomography (USST) has shown great prospects for breast cancer diagnosis. Its reconstruction is a complex nonlinear mapping from the projection data to the sound-speed image (SSI). The traditional reconstruction methods mainly include the ray-based methods and the waveform-based methods. The ray-based methods with linear approximation have low computational cost but low reconstruction quality; the full-wave-based methods with the complex nonlinear model have high quality but high cost. To achieve both high quality and low cost, we introduced traditional linear approximation as prior knowledge into a deep neural network and treated the complex nonlinear mapping of USST reconstruction as a combination of linear mapping and nonlinear mapping. The linear mapping was seamlessly implemented with a fully connected layer and initialized using Tikhonov pseudo-inverse matrix. The nonlinear mapping was implemented using a U-shape Net (U-Net). Furthermore, we proposed the Tikhonov U-shape Net (TU-Net), in which the linear mapping was done before the nonlinear mapping, and the U-shape Tikhonov Net (UT-Net), in which the nonlinear mapping was done before the linear mapping. Both numerical simulation and prototype experiment were conducted. In the simulation, the root-mean-square error was 5.12 and 4.13 m/s for the UT-Net and the TU-Net, the peak signal-to-noise ratio was 51.20 and 53.33 dB, the structural similarity was 0.9647 and 0.9775, and the reconstruction time was 5.0 and 4.9 ms, respectively. In the experiment, the SSIs obtained by the proposed methods exhibited high sound-speed accuracy. Both the UT-Net and the TU-Net achieved high quality and low computational cost.

Indexed by:Journal paper

Translation or Not:no

Date of Publication:2022-07-31

Included Journals:SCI

Attachments: