Model-based T1, T2* and Proton Density Mapping Using a Bayesian Approach with Parameter Estimation and Complementary Undersampling Patterns

TL;DR

This paper introduces a Bayesian reconstruction method called AMP-PE for quantitative MRI that automatically estimates hyperparameters and optimally uses undersampling patterns, improving accuracy over existing approaches.

Contribution

The paper presents a novel nonlinear message passing framework for joint parameter and image recovery, incorporating complementary undersampling patterns for optimal quantitative MRI reconstruction.

Findings

AMP-PE outperforms $l_1$-norm minimization and existing joint-recovery methods.

Using identical k-space sampling patterns across echo times improves T1 mapping.

Complementary sampling patterns enhance T2* and proton density mapping.

Abstract

Purpose: To achieve automatic hyperparameter estimation for the joint recovery of quantitative MR images, we propose a Bayesian formulation of the reconstruction problem that incorporates the signal model. Additionally, we investigate the use of complementary undersampling patterns to determine optimal undersampling schemes for quantitative MRI. Theory: We introduce a novel nonlinear approximate message passing framework, referred to as ``AMP-PE'', that enables the simultaneous recovery of distribution parameters and quantitative maps. Methods: We employed the variable flip angle multi-echo (VFA-ME) method to acquire measurements. Both retrospective and prospective undersampling approaches were utilized to obtain Fourier measurements using variable-density and Poisson-disk patterns. Furthermore, we extensively explored various undersampling schemes, incorporating complementary patterns across different flip angles and/or echo times. Results: AMP-PE adopts a model-based joint recovery strategy, it outperforms the $l_1$-norm minimization approach that follows a decoupled recovery strategy. A comparison with an existing joint-recovery approach further demonstrates the advantageous outcomes of AMP-PE. For quantitative $T_1$ mapping using VFA-ME, employing identical k-space sampling patterns across different echo times produced the best performance. Whereas for $T_2^*$ and proton density mappings, using complementary sampling patterns across different flip angles yielded the best performance. Conclusion: AMP-PE is equipped with built-in parameter estimation, and works naturally in clinical settings with varying acquisition protocols and scanners. It also achieves improved performance by combining information from the MR signal model and the sparse prior on images.