Résumé
Validation of Magnetic Resonance Elastography (MRE) often relies on simple geometric phantoms which lack the complexity and heterogeneity of biological tissues. This limitation creates a methodological gap when assessing reconstruction algorithms intended for heterogeneous clinical environments. The aims were to demonstrate the feasibility of fabricating advanced polyphasic plastisol phantoms using a welding method and to evaluate reconstruction performance for phantoms with increasing geometric complexity specifically at material interfaces. Four plastisol phantoms (32 mm diameter, 17 mm height) were developed: homogeneous, inclusions, sectorial divisions, and a 3×3 checkerboard grid. Acquisitions were performed at 9.4T using a fast spin-echo sequence with sinusoidal motion-encoding gradients (6 G/cm) and eight temporal phase offsets, an isotropic 0.8mm resolution, a 64×64 matrix, and T2 mapping. Shear stiffness (µ) and damping ratio (ξ) were reconstructed using Algebraic Inversion of the Differential Equation (AIDE) and Non-Linear Inversion (NLI). Performance was evaluated comparing welded samples to non-welded controls and analyzing property deviations from the homogeneous case. Polyphasic samples demonstrated mechanical properties identical to homogeneous (µ = 9.9 ± 0.5 vs. 10.1 ± 0.5 kPa), confirming that the thermal process preserves the material's intrinsic properties. The resulting interfaces exhibit full mechanical continuity, acting as a single block. While both algorithms successfully captured structural heterogeneities, increasing geometric complexity induced systematic biases, particularly the underestimation of stiff regions and the overestimation of damping ratios near boundaries. Plastisol thermal fusion assembly enables the fabrication of stable, complex phantoms with precise mechanical control. Although the heterogeneities in this study were primarily 2D due to the ratio between shear wavelength and sample dimensions, the process is inherently compatible with 3D voxel-wise assembly. While formal tensile testing was not performed, basic stress tests confirmed that the interfacial cohesion is largely sufficient for the low-amplitude strains involved in MRE. These phantoms provide a versatile experimental framework for identifying algorithmic limitations at tissue interfaces and benchmarking advanced reconstruction methods in biologically relevant models.