Amplitude Based Segmentation of Ultrasound Echoes for Attenuation Coefficient Estimation

In vivo ultrasound attenuation coefficient measurements are of interest as they can provide insight into tissue pathology. They are also needed so that measurements of the tissue's frequency dependent ultrasound backscattering coefficient may be corrected for attenuation. In vivo measurements of the attenuation coefficient are challenging because it has to be estimated from the depth dependent decay of backscatter signals that display a large degree of magnitude variation. In this study we describe and evaluate an improved backscatter method to estimate ultrasound attenuation which is tolerant to the presence of some backscatter inhomogeneity. This employs an automated algorithm to segment and remove atypically strong echoes to lessen the potential bias these may introduce on the attenuation coefficient estimates. The benefit of the algorithm was evaluated by measuring the frequency dependent attenuation coefficient of a gelatine phantom containing randomly distributed cellulose scatterers as a homogeneous backscattering component and planar pieces of cooked leek to provide backscattering inhomogeneities. In the phantom the segmentation algorithm was found to improve the accuracy and precision of attenuation coefficient estimates by up to 80% and 90%, respectively. The effect of the algorithm was then measured invivo using 32 radiofrequency B-mode datasets from the breasts of two healthy female volunteers, producing a 5 to 25% reduction in mean attenuation coefficient estimates and a 30 to 50% reduction in standard deviation of attenuation coefficient across different positions within each breast. The results suggest that the segmentation algorithm may improve the accuracy and precision of attenuation coefficient estimates invivo.