Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Constrained and Regularized Quantitative Ultrasound Parameter Estimation using ADMM

Ali K. Z. Tehrani, Hassan Rivaz, Ivan M. Rosado-Mendez

TL;DR

This work addresses nonphysical estimates in spectral QUS parameter estimation by incorporating minimum feasible value constraints into an ADMM framework, yielding Constrained-ADMM (C-ADMM). The method preserves the established regularization scheme, using $L2$ for attenuation and $L1$ for backscatter and frequency-dependence, while enforcing physical bounds via a minimum parameter vector $\beta$ and a clipping mechanism with a latent variable. Validation on a Gammex phantom shows that C-ADMM eliminates negative attenuation estimates and reduces bias and variance in background QUS parameters, especially in regions affected by strong specular reflectors, with negligible additional computation. The approach enhances robustness and reliability of QUS tissue characterization, with potential applicability to broader QUS estimation problems and clinical imaging scenarios.

Abstract

Regularized estimation of quantitative ultrasound (QUS) parameters, such as attenuation and backscatter coefficients, has gained research interest. Recently, the alternating direction method of multipliers (ADMM) has been applied successfully to estimate these parameters, by utilizing L2 and L1 norms for attenuation and backscatter coefficient regularization, respectively. While this method improves upon previous approaches, it does not fully leverage the prior knowledge of minimum physically feasible parameter values, sometimes yielding values outside the realistic range. This work addresses this limitation by incorporating minimum QUS parameter values as constraints to enhance ADMM estimation. The proposed method is validated using experimental phantom data.

Constrained and Regularized Quantitative Ultrasound Parameter Estimation using ADMM

TL;DR

This work addresses nonphysical estimates in spectral QUS parameter estimation by incorporating minimum feasible value constraints into an ADMM framework, yielding Constrained-ADMM (C-ADMM). The method preserves the established regularization scheme, using for attenuation and for backscatter and frequency-dependence, while enforcing physical bounds via a minimum parameter vector and a clipping mechanism with a latent variable. Validation on a Gammex phantom shows that C-ADMM eliminates negative attenuation estimates and reduces bias and variance in background QUS parameters, especially in regions affected by strong specular reflectors, with negligible additional computation. The approach enhances robustness and reliability of QUS tissue characterization, with potential applicability to broader QUS estimation problems and clinical imaging scenarios.

Abstract

Regularized estimation of quantitative ultrasound (QUS) parameters, such as attenuation and backscatter coefficients, has gained research interest. Recently, the alternating direction method of multipliers (ADMM) has been applied successfully to estimate these parameters, by utilizing L2 and L1 norms for attenuation and backscatter coefficient regularization, respectively. While this method improves upon previous approaches, it does not fully leverage the prior knowledge of minimum physically feasible parameter values, sometimes yielding values outside the realistic range. This work addresses this limitation by incorporating minimum QUS parameter values as constraints to enhance ADMM estimation. The proposed method is validated using experimental phantom data.
Paper Structure (10 sections, 10 equations, 3 figures, 1 table)

This paper contains 10 sections, 10 equations, 3 figures, 1 table.

Table of Contents

  1. Introduction
  2. Material and Methods
  3. Problem formulation
  4. ADMM method
  5. Proposed method
  6. Datasets
  7. Quantitative metrics
  8. Results
  9. Discussion
  10. Conclusion

Figures (3)

  • Figure 1: The B-mode image of the phantom.
  • Figure 2: QUS parameter estimation results. $\alpha_t:$ Attenuation Coefficient (A), $b_t:$ Backscattering Coefficient in dB (B), and $n_t:$ frequency dependency (C). The marked area with red arrows illustrates the region where there are artifacts.
  • Figure 3: Bias error analysis of the quantitative ultrasound (QUS) parameters. (a) Attenuation coefficient bias ($\alpha_t$), (b) backscattering coefficient bias ($b_t$), and (c) frequency–dependence power bias ($n_t$).