Gradient-Based Optimization for Poroelastic and Viscoelastic MR Elastography.

Area of Science:

• Medical Imaging
• Computational Physics
• Biomedical Engineering

Background:

• Magnetic Resonance Elastography (MRE) is crucial for non-invasively assessing tissue stiffness.
• Solving inverse problems in MRE requires efficient gradient computation for accurate material property reconstruction.
• Traditional adjoint methods necessitate specific matrix properties, limiting their applicability.

Purpose of the Study:

• To develop and validate an efficient gradient computation algorithm for MRE inverse problems.
• To demonstrate that the proposed method does not require the self-adjoint property of stiffness matrices.
• To achieve computational efficiency comparable to classic adjoint methods, independent of parameter count.

Main Methods:

• A generalized adjoint method based on Lagrangian formulation was developed.
• The algorithm was implemented for material property reconstruction using poroelastic and viscoelastic models.
• Gradient and Hessian-based optimization techniques were applied to simulation, phantom, and in vivo brain data.

Main Results:

• The proposed gradient computation scheme is feasible and efficient for MRE inverse problems.
• The method achieves computational performance comparable to classic adjoint methods.
• It requires only two forward solutions, irrespective of the number of estimated parameters.

Conclusions:

• The novel gradient computation algorithm enhances the efficiency of MRE inverse problem solving.
• This method offers a more flexible and computationally advantageous approach for material property reconstruction in MRE.
• The findings support the broader application of advanced MRE techniques in clinical and research settings.