The Effects of Noise and Spatial Sampling on Identification of Material Parameters by Magnetic Resonance Elastography

Abstract

In-vivo measurement of the mechanical properties of soft tissues is challenging but may provide invaluable data in biomechanics and medicine. Different experimental techniques are currently used to measure 3D strain or displacements fields under static or dynamic loading. Different analysis methods have also been developed to extract the materials mechanical properties from these data (finite element model + cost function, local fitting of the propagation equation, etc). The main difficulty of these extractions consists in dealing with the experimental noise and spatial derivatives required during the processing. The aim of this communication is to provide an insight into the Optimized Virtual Field Method (OVFM, Avril et al. Comput Mech 34:439–452, 2004) abilities to identify mechanical properties while analyzing noisy data. MRE data over harmonically loaded soft materials has been simulated. The effect of noise and spatial sampling has then been studied while identifying locally a viscoelastic model with the OVFM. This noise effect study provided an a priori criterion to estimate the local identification quality. The results also helped to underline and estimate identification biases induced by the noise and spatial subsampling. A viscoelastic model has then been identified over a real experimental data set, providing 3D mechanical parameters maps.