Fast L1-Regularized QSM Reconstruction via Variable Splitting
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current MRI techniques for quantitative susceptibility mapping are inefficient, requiring long reconstruction times and increased patient discomfort due to the need for multiple orientations and regularization methods, which complicate the estimation of magnetic susceptibility and iron concentration in tissues.
Innovation Solution
The method employs a fast l1-regularized quantitative susceptibility mapping algorithm using variable splitting and regularization parameters to iteratively reconstruct susceptibility maps, incorporating phase unwrapping and background phase removal, which significantly reduces reconstruction time and improves accuracy.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If conventional regularization methods are used for susceptibility inversion, then measurement precision is improved, but loss of time increases due to long reconstruction times
Solution Approach 1:
The patent changes the regularization parameter from l2-norm to l1-norm in the susceptibility inversion process. This parameter change enables sparsity promotion in the susceptibility map while maintaining accuracy, and when combined with variable splitting, allows for faster convergence and reduced reconstruction time without sacrificing measurement precision
Solution Approach 2:
The patent applies variable splitting to divide the complex susceptibility inversion problem into separate sub-problems that can be solved iteratively. By segmenting the inversion process into manageable steps with alternating optimization of different variables, the method achieves both high precision and computational efficiency
2Reliability
If multi-orientation acquisition is used to stabilize susceptibility reconstruction, then reliability is improved, but loss of time increases due to increased scan time
Solution Approach 1:
The patent performs preliminary action by applying l1-regularization and variable splitting to the single-orientation data before final reconstruction. This pre-processing and regularization approach stabilizes the inversion process upfront, eliminating the need for time-consuming multi-orientation acquisitions while maintaining reconstruction reliability
3Productivity
If l1-regularization with variable splitting is used, then productivity is improved through faster reconstruction, but device complexity increases due to iterative algorithms
Solution Approach 1:
The patent substitutes traditional mechanical iterative reconstruction methods with an optimized l1-regularized variable splitting algorithm. This algorithmic substitution uses mathematical optimization techniques that converge faster than conventional methods, achieving high productivity despite increased computational complexity through efficient implementation
Applied Scientific Principles
This section explains which scientific principles are used to turn an abstract innovation direction into a practical engineering solution.
Function Achieved in This Case
This approach achieves a twenty-fold speed-up in reconstruction time compared to conventional methods, enabling clinically reasonable processing of large datasets and facilitating the investigation of neurodegenerative diseases by providing high-resolution susceptibility maps in under a minute.
Implementation Method 1
acquire data from a subject using a magnetic resonance imaging ('MRI') system
Data Source
AI summary
Described here are systems and methods for quantitative susceptibility mapping (“QSM”) using magnetic resonance imaging (“MRI”). Susceptibility maps are reconstructed from phase images using an automatic regularization technique based in part on variable splitting. Two different regularization parameters are used, one, λ, that controls the smoothness of the final susceptibility map and one, μ, that controls the convergence speed of the reconstruction. For instance, the regularization parameters can be determined using an L-curve heuristic to find the parameters that yield the maximum curvature on the L-curve. The μ parameter can be determined based on an l2-regularization and the λ parameter can be determined based on the iterative l1-regularization used to reconstruct the susceptibility map.