Ultrasound elastography performance enhancement using wavelet denoising

Abstract

Ultrasound elastography is a kind of technique of obtaining the tissue relative stiffness information, which plays an important role in early tumor diagnosis. The increases in overlap between the data segments on the order of 90% or higher are essential to improve axial resolution in elastogaphy. However, correlated errors in displacement estimates increase dramatically with the increase in the overlap, and generate the so-called “worm” artefact. This results in a trade-off between the SNRe(CNRe) and the axial resolution. This paper presents a wavelet shrinkage denoising in strain estimates to reduce the worm artifact at a high overlap. The soft threshold function is selected and every adaptive threshold selection rule is tested to denoise the noisy strain estimates. The simulation results illustrate that the presented technique can efficiently denoise worm artefact and enhance the elastogram performance indices such as elastographic SNRe, CNRe while maintaining the high axial resolution. Elastogram obtained by wavelet denoising has the closest correspondence with ideal strain image(high correlation coefficient). In addition, the results also demonstrate that wavelet shrinkage denoising applied in strain estimates can obtain better image quality parameters than that applied in displacement estimates. The phantom experiments also show the similar elastogram performance improvement by using the proposed denoising technique.