Level Set Diffusion for MRE Image Enhancement

摘要

Magnetic resonance elastography (MRE) is an emerging technique for noninvasive imaging of tissue elasticity. Proprietary algorithms are used to reconstruct tissue elasticity from the images of wave propagation within soft tissue. Elasticity reconstruction suffers from interfering noise and outliers. The interference causes biased elasticity and undesired artifacts in the reconstructed elasticity map. Anisotropic geometric diffusion is able to suppress image noise while enhance inherent features. Therefore we integrate anisotropic diffusion with level set methods for numerical enhancement of MRE wave images. Performance evaluation of the proposed level set diffusion (LSD) approach was conducted on both synthetic and real MRE datasets. Experimental results confirm the effectiveness of LSD for MRE image enhancement and direct inversion.