Attenuation Compensation in Ultrasound Images

[paper]

Ultrasound B-scan exhibits shadowing and enhancement artifacts due to acoustic wave propagation and spatially varying attenuation across tissue layers. These artifacts hinder visualization, interpretation and quantitative pathological analysis of clinical features for diagnosis. We have developed a novel image-based approach to generate high quality attenuation estimation and compensation, along with automatically delineated structure boundaries.

Obtaining an estimate of the original backscatter and attenuation given only the observed image with unknown background noise is an underdetermined, inverse problem. We formulate a mathematical model for jointly estimating the backscatter (b), contour (C), and attenuation (a) fields of an object by minimizing a cost functional with region-based isotropic regularizations. A three-step BCA alternating minimization procedure is adopted towards a tractable solution. This leads to solving equivalent second order Euler-Lagrange equations for b, a mixed parabolic integral equation for a, and a curve evolution equation for C . We have proved analytically the existence, uniqueness, convergence and stability of the solutions.

The BCA method demonstrates a higher artifact suppression capacity on simulated and in vivo ultrasound images. The extracted backscatter has significantly lower and more uniform speckle. It works equally well on data with a fairly complex attenuation field.