A novel denoising strategy for ultrasound elastography

Abstract

Ultrasound imaging, while offering real-time tissue monitoring, portability, and other benefits, often suffers from poor signal to noise ratio (SNR). Noise has a particularly detrimental effect in US elastography since it gets amplified during the elastic modulus reconstruction process due to computation of high-order spatial derivatives of the deformation field needed to calculate the shear wave velocity within the sample. In this work, we introduce a novel scheme to denoise deformation fields using physically motivated “compatibility conditions,” which are imposed as constraints. These conditions, well established in the solid mechanics literature, ensure that the deformation field remains continuous without exhibiting cracks, folds, or slips. Spatially uncorrelated noise gets significantly reduced when we apply them. Importantly, this approach avoids unnecessary smoothing and the ensuing loss of accuracy commonly observed in other denoising methods. We demonstrate the efficacy of this method using planar deformation fields obtained from simulations and those measured on an elasticity quality assurance (QA) phantom using acoustic radiation force impulse imaging (ARFI). Our pipeline results in smoother elasticity maps with 2x improved SNR, confirming the effectiveness of the approach.