A computer simulation study of soft tissue characterization using low-frequency ultrasonic tomography

•We explore the potential of low-frequency tomography for breast cancer diagnosis.•Using full data, reflection and transmission, allows sub-wavelength reconstruction.•The problem is treated as a coefficient inverse problem for the wave equation.•The algorithms are based on direct computation of the gradient of the residual.•Parallel algorithms focus on the use of GPU clusters and supercomputers. We investigate the potential of using ultrasonic diffraction tomography technique for characterization of biological tissues. Unlike most of other studies where ultrasonic tomography operates at frequencies higher than 1MHz, low-frequency tomography uses lower frequencies on the order of 0.3–0.5MHz. Such a choice is due to low attenuation at these frequencies, resulting in higher precision of input data. In this paper we explore transmission and reflection schemes for both 2D (layer-by-layer) and 3D tomography. We treat inverse tomography problems as coefficient inverse problems for the wave equation. The time-domain algorithms employed for solving the inverse problem of low-frequency tomography focus on the use of GPU clusters. The results obtained show that a spatial resolution of about 2–3mm can be achieved when operating at the wavelength of about 5mm even using a stationary 3D scheme with a few fixed sources and no rotating elements. The study primarily focuses on determining the performance limits of ultrasonic tomography devices currently designed for breast cancer diagnosis.