Geometric a priori informed bent-ray tracing for accelerated sound speed imaging in ultrasound computed tomography

Bent-ray tracing ultrasound computed tomography (USCT) is a promising technique for breast cancer screening which quantitatively provides speed-of-sound (SOS) distribution in human breasts. In this modality, SOS images are reconstructed with an iterative process to match the measured time-of-flights and the ones predicted by Eikonal equation solved with the fast marching method (FMM). The Eikonal equation is meant to be applied in SOS heterogeneous media and its evaluation with FMM is an computational expensive process. However, in USCT, the object is placed in a homogeneous coupling medium. Thus, the acoustic environment is formed by two parts, the homogeneous background (coupling medium) and the heterogeneous object. In this work, we leverage this strong a priori information and propose a method to accelerate SOS image formation for bent-ray tracing USCT. We show that, given the boundary information of the object, Eikonal equation only needs to be evaluated in a limited area covering the object. For that, the partial FMM and the associated ray-tracing strategy are proposed to reduce the computational cost of the forward modeling. We also managed to restrict image reconstruction area inside the object for improved convergence rate of the optimization. Both the simulation and phantom imaging experiments with ring transducer arrays demonstrated that the proposed method reduces the reconstruction time in an object size dependent manner. For the object occupying 20.3% to 56.3% of the image field of the ring array, we observed 30.1%-61.9% reduction in image reconstruction time without sacrificing the image quality, compared to classical method. The proposed strategy can be adopted for fast SOS imaging with bent-ray tracing USCT to improve patient throughput for breast cancer screening.