[Paper Review] Stability Analysis for Magnetic Resonance Elastography

Habib Ammari, Alden Waters+1 more|arXiv (Cornell University)|Sep 17, 2014

Ultrasound Imaging and ElastographyMedicine22 references3 citations

TL;DR

This paper establishes Lipschitz stability estimates for reconstructing shear modulus distributions in magnetic resonance elastography (MRE) from internal displacement measurements using the Stokes system as a model for quasi-incompressible tissues. By reducing the elasticity equations to the Stokes system in the limit of high compressional modulus, the authors prove that the inverse problem is well-posed and demonstrate convergence of the Landweber iteration scheme for shear modulus reconstruction in both two and three dimensions, with additional measurements required in 3D to ensure stability.

ABSTRACT

We consider the inverse problem of finding unknown elastic parameters from internal measurements of displacement fields for tissues. The measurements are made on the entirety of a smooth domain. Since tissues can be modeled as quasi-incompressible fluids, we examine the Stokes system and consider only the recovery of shear modulus distributions. Our main result is to establish Lipschitz stable estimates on the shear modulus distributions from internal measurements of displacement fields. These estimates imply convergence of a numerical scheme known as the Landweber iteration scheme for reconstructing the shear modulus distributions.

Motivation & Objective

• To address the inverse problem of recovering shear modulus distributions from internal displacement measurements in magnetic resonance elastography (MRE).
• To establish mathematical stability estimates for the reconstruction of shear modulus from measured displacement fields in soft tissues.
• To prove convergence of the Landweber iteration scheme for reconstructing shear modulus under appropriate regularity and stability conditions.
• To extend the analysis to three dimensions, where additional internal measurements are required for stability.
• To provide a theoretical foundation for numerical reconstruction methods in MRE using over-determined systems of PDEs.

Proposed method

• Reduce the time-harmonic elasticity system to the Stokes system in the limit of infinite compressional modulus (λ → ∞), modeling quasi-incompressible tissues.
• Formulate the inverse problem as an over-determined system of partial differential equations for the shear modulus μ.
• Prove that the resulting system is elliptic and satisfies the Lopatinskii boundary condition, ensuring well-posedness.
• Use stability estimates for elliptic systems to derive Lipschitz bounds on the shear modulus from internal displacement data.
• Introduce a discrepancy functional J[μ] based on the L² norm of the difference between computed and measured displacement fields.
• Apply the Landweber iteration scheme via gradient descent on the functional, with adjoint system for computing the Fréchet derivative.

Experimental results

Research questions

• RQ1 Can Lipschitz stability estimates be established for reconstructing shear modulus from internal displacement measurements in MRE?
• RQ2 How does the inverse problem for shear modulus behave under the Stokes system approximation in quasi-incompressible tissues?
• RQ3 What conditions ensure convergence of the Landweber iteration scheme in two and three dimensions?
• RQ4 Why are additional internal measurements necessary in three-dimensional MRE for stable reconstruction?
• RQ5 Can the kernel of the over-determined system be removed using multifrequency measurements or other enhancements?

Key findings

• Lipschitz stability estimates are established for the reconstruction of shear modulus distributions from internal displacement data in both two and three dimensions.
• In three dimensions, the addition of a second displacement measurement under different boundary conditions is required to ensure stability and convergence.
• The Landweber iteration scheme converges in the H⁴(Ω) norm to the true shear modulus distribution μ_tr when the initial guess is sufficiently close.
• The convergence result holds under the assumption that the kernel K of the over-determined system (L, B) is trivial.
• The theoretical framework supports the use of gradient-based numerical schemes for MRE reconstruction with guaranteed convergence under regularity and stability conditions.
• The analysis provides a rigorous mathematical basis for numerical reconstruction in MRE, particularly for early cancer detection via high-resolution shear modulus imaging.