Maximum Likelihood MRF Reconstruction for MRI Parameter Maps
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional magnetic resonance fingerprinting (MRF) reconstruction methods are computationally inefficient and statistically suboptimal, leading to biased parameter estimation and dependency on long acquisition sequences, especially for parameters like T2 maps, and do not fully exploit the signal-to-noise ratio (SNR) benefits offered by phased array coils.
Innovation Solution
A maximum likelihood (ML) MRF reconstruction framework is introduced, which performs direct estimation of multiple parameter maps from highly undersampled k-space data, incorporating variable splitting, alternating direction method of multipliers (ADMM), and variable projection (VARPRO) to optimize the reconstruction process, reducing acquisition time while maintaining accuracy.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If conventional template-matching MRF reconstruction is used, then the reconstruction process is simple and robust, but it is computationally inefficient and statistically suboptimal leading to biased parameter estimation
Solution Approach 1:
The patent transforms the reconstruction problem from simple template matching to a maximum likelihood estimation framework by changing the statistical parameters and optimization criteria. This involves formulating the reconstruction as an optimization problem that maximizes the likelihood function based on the signal model and noise characteristics, thereby achieving statistically optimal parameter estimation while maintaining computational tractability through efficient optimization algorithms.
Solution Approach 2:
The patent replaces the mechanical template-matching procedure with a statistical optimization approach. Instead of simply comparing acquired signals to pre-computed templates, the system uses maximum likelihood estimation with optimized algorithms (such as alternating direction method of multipliers and variable projection) to iteratively refine parameter estimates, substituting the brute-force mechanical matching with a more efficient statistical framework.
2Measurement precision
If long acquisition sequences are used to improve T2 map accuracy, then parameter estimation accuracy improves, but acquisition time increases
Solution Approach 1:
The patent implements an iterative reconstruction framework where the maximum likelihood estimation process continuously refines parameter estimates based on feedback from the optimization algorithm. The alternating direction method of multipliers and variable projection techniques provide feedback mechanisms that progressively improve T2 map accuracy without requiring proportionally longer acquisition times, as the iterative process efficiently extracts information from the available data.
Solution Approach 2:
The patent uses preliminary action by pre-computing signal evolutions from the Bloch equation to create a dictionary of templates that encode the relationship between acquisition parameters and signal responses. This preliminary preparation enables the maximum likelihood estimation to efficiently compare acquired signals against pre-characterized tissue behaviors, reducing the need for extended acquisition sequences while maintaining T2 mapping accuracy.
3Loss of time
If highly undersampled k-space data is used, then acquisition time is reduced, but reconstruction quality decreases and artifacts increase
Solution Approach 1:
The patent changes the reconstruction parameters from simple gridding or Fourier-based methods to maximum likelihood estimation with regularized optimization. By formulating the reconstruction as a statistical optimization problem that incorporates the signal model and noise characteristics, the system can effectively utilize highly undersampled k-space data while suppressing artifacts and maintaining reconstruction quality through optimal parameter estimation.
Solution Approach 2:
The patent replaces conventional mechanical reconstruction methods (such as direct Fourier transform or simple gridding) with a statistical optimization approach. The maximum likelihood estimation framework, implemented through algorithms like alternating direction method of multipliers and variable projection, substitutes the straightforward but artifact-prone mechanical reconstruction with a more sophisticated statistical framework that can recover high-quality images from highly undersampled data by optimally utilizing the available signal information.
Data Source
AI summary
Systems and methods for reconstructing magnetic resonance (MR) tissue parameter maps of a subject from magnetic resonance fingerprinting (MRF) data acquired using a magnetic resonance imaging (MRI) system. The method includes providing MRF data acquired from a subject using an MRI system and performing an iterative, maximum-likelihood reconstruction of the MRF data to create MR tissue parameter maps of the subject.