Strategies to Improve the Performance of USCT Algorithms

Abstract

In recent years, several studies about the generation of ultrasound tomographic images (USCT) have been proposed. The initialization of the medium to be reconstructed has a great impact on the final result of some of those algorithms. Taking advantage of the reflection and transmission phenomenon, interesting results have been obtained which extract information from the propagation medium. However objective studies about the effects of initialization in USCT reconstruction are lacking. To contribute to this topic, our work presents, on simulated medium, a comparison of the performance of a USCT algorithm with two types of initialization, one with traditional initialization (homogeneous medium) and other with information coming from other techniques. In this study, the following were used: Distorted Born Iterative Method (DBIM) as the USCT-algorithm; a Matlab Toolbox (k-wave) for the simulation; synthetic transmission aperture as edge information extractor; Reflection mask to limit the reconstruction area; Simultaneous Algebraic Reconstruction Technique for transmission tomography algorithm. Two sets of data were generated: the first with speed of sound between [1400–1680] m/s and the second with speed of sound between [1250–1830] m/s simulating a higher contrast medium; Normalized-Root-Mean-Square-Error (NRMSE) and the Structural similarity (SSIM or Q) were utilized for the evaluation. The investigation, based on simulations, suggests that the use of strategies that delimit or decrease the number of variables to be found (mask) and that applying information from reflection and transmission phenomenon to USCT-algorithms can improve its performance, offering better reconstructions and consequently more accurate information of the medium.