【摘要】 磁共振弹性成像(magnetic resonance elastography, MRE)作为一种无创定量测量软组织力学特性的成像技术，已在肝纤维化分期和神经退行性疾病诊断等领域展现出广阔的应用前景。然而，从实测剪切波位移场中准确反演组织剪切模量面临病态逆问题、高阶微分噪声放大与组织异质性失配等挑战。本文首先系统回顾了基于解析假设的直接反演、相位梯度及局部频率估计等快速算法及其改进；总结了有限元迭代、子区域分解与无散度约束等数值优化方法在复杂生物力学建模中的应用；随后探讨基于有限元仿真的神经网络和基于物理模型驱动行波展开的神经网络反演新范式；最后评述当前算法在黏性参数恢复、多尺度异质性处理与端到端联合优化方面的机遇与瓶颈。本文通过综述最新进展与发展趋势，旨在为MRE反演算法的性能提升与临床转化提供理论依据和技术指导。更多还原

【Abstract】 Magnetic resonance elastography(MRE) has emerged as a powerful non-invasive technique for quantifying soft tissue mechanics, with demonstrated utility in liver fibrosis staging and neurodegenerative disease assessment. However, solving the inverse problem of reconstructing shear modulus from measured displacement fields remains ill-posed, with challenges including noise amplification by high-order spatial derivatives and model mismatch in heterogeneous tissues. This review first surveys analytical inversion approaches(direct inversion, phase gradient, and local frequency estimation), and their enhancements for rapid modulus mapping. Then numerical optimization methods are summarized, including finite element-based nonlinear inversion, sub-zone decomposition, and divergence-free constraints, which enable complex biomechanical modeling. Next, emerging data-driven paradigms with finite element method and highlighting neural network inversion that integrates traveling wave expansion with physics-based consistency are examined. Finally, future opportunities and obstacles in viscoelastic parameter recovery, multiscale heterogeneity, and end-to-end joint optimization are discussed. By synthesizing recent advances and outlining remaining challenges, this review aims to guide the development of more robust and clinically translatable MRE inversion algorithms.更多还原