LAUSR

BACKGROUND AND OBJECTIVE The resolvable scatterer spacing related to biological tissue microstructures is a quantitative signature used for the disease diagnosis and tissue classification. In the present study, a method by locating optimally matched Gabor atoms (LOMGA) from ultrasound RF echo signals is proposed to improve the inter-scatterer spacing (ISS) estimation. METHOD A series of Gabor atoms are obtained from the signals with a matching pursuit algorithm. Then, the optimum atoms highly correlated with the coherent components are automatically selected according to the second-order difference of the reconstructed signal-to-residue ratio. The distances between the locations of adjacent atoms are applied to estimate the ISSs. In the simulation experiments, four regular degrees of the scatterer distributions are modeled with the Gamma distribution. One hundred sets of ultrasound RF echo signals are simulated based on the regular and diffuse scatterer distributions, and then combined to generate signals with preset coherent-to-diffuse ratios (CDRs). The accuracy performance of the LOMGA method is compared with that based on wavelet transform (WT) algorithm. In the microwave ablation experiments, the ultrasound RF echo signals of the region of interest (ROI) are collected from the normal and coagulated porcine liver tissues. The means and standard deviations of the LOMGA-based ISSs are compared with the WT-based results. RESULTS The results based on simulated signals with CDRs from 10 dB to -10 dB demonstrate that the proposed method improves the estimation accuracies of the mean ISSs by 5.10%, 9.00%, 19.80%, and 23.82%, and reduces the mean standard deviations by 27.20%, 22.50%, 11.50%, and 4.49% more than the WT method for the four regularities, respectively. The performance of the LOMGA method is also verified with the ultrasound RF echo signals from ex vivo porcine liver tissues in microwave ablation experiments. CONCLUSIONS It is concluded that the LOMGA method can provide more accurate and stable ISS estimation, which improves the performance of the tissue characterization with ISS-based quantitative ultrasound.