Artifact suppression in freehand ultrasound elastography using Multiscale Principal Component Analysis

摘要

Ultrasound elastography is a noninvasive technique for mapping the elasticity of soft tissues as an image called elastogram (axial-strain image). Applying a small axial pressure on the tissue surface is the main requirement of ultrasound elastography. If this pressure is applied manually by the ultrasound probe the technique is called “freehand ultrasound elastography”. Non-ideal manual compressions lead to emergence of undesired artifacts in the elastograms which degrade their quality and restrict their clinical applicability. In this paper we propose a method based on Multiscale Principal Component Analysis (MSPCA) to suppress the elastographic artifacts and yield refined elastograms with better Elastographic Signal to Noise Ratio (SNRe) and Elastographic Contrast to Noise Ratio (CNRe). We applied our proposed method to a freehand elastographic dataset of a phantom which mimicked a hard tumor in a soft background tissue. The results showed significant improvements in average SNRe and CNRe as evaluated by paired sample t-test (p<;0.001). Also, it was shown the proper selection of mother wavelet and the number of wavelet decomposition levels is very important to prevent the emergence of undesired patterns called Zipper-like artifacts in the refined elastograms.