Voxel-Based Tissue Parameterization with Super-Resolution Quantitative MRI
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing methods for obtaining tissue parameters, such as brain tissue, face challenges in achieving high spatial resolution, high signal-to-noise ratio (SNR), and short acquisition time, with diffusion tensor models being too simplistic and requiring complex calculations that are infeasible for high-resolution imaging.
Innovation Solution
A combined super-resolution imaging and quantitative MRI modeling technique is applied to transform weighted MRI volume scans into a parameterized voxel-based model, using iterative loops to optimize parameters per voxel, avoiding complex derivative calculations and constraints.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Loss of time
If diffusion tensor model is used for super-resolution reconstruction, then acquisition time is reduced and SNR is improved, but the model is too simple to accurately describe underlying microstructure and cannot model multiple fibre orientations
Solution Approach 1:
The patent changes the parameterization approach from simple diffusion tensor model to a more comprehensive model that includes multiple fibre orientations and multiple microscopic compartments. This allows accurate description of complex tissue microstructure while maintaining the super-resolution reconstruction framework that reduces acquisition time.
Solution Approach 2:
The patent uses a composite modeling approach combining multiple compartments (e.g., intra-axonal, extra-axonal, CSF) within a single voxel model. This composite structure enables the model to capture multiple fibre orientations and different tissue microenvironments simultaneously, resolving the limitation of the simple diffusion tensor model.
2Measurement precision
If more parameters are calculated for diffusion tensor model, then Hessian matrix becomes infeasible to store and calculate, but more parameters are needed for accurate tissue characterization
Solution Approach 1:
The patent segments the parameter estimation process into voxel-wise independent calculations. By solving the optimization problem separately for each voxel rather than globally, the computational complexity is dramatically reduced, making it feasible to calculate multiple parameters without requiring storage and manipulation of large Hessian matrices.
Solution Approach 2:
The patent applies partial action by performing computations only where needed (voxel-wise rather than globally) and using appropriate approximation methods for each voxel. This reduces the overall computational burden while maintaining sufficient accuracy for tissue characterization.
3Measurement precision
If linear inequality constraints are imposed in spherical deconvolution, then multi-compartment modeling is achieved, but optimization problem converges too slowly for high resolution images
Solution Approach 1:
The patent performs preliminary actions by pre-defining the compartment structure and basis functions before the optimization process. This preparation allows the subsequent optimization to converge faster, as the search space is already structured and constrained in a physically meaningful way, avoiding slow convergence even with linear inequality constraints.
Solution Approach 2:
The patent changes the optimization approach by reformulating the problem to work with a reduced parameter set that is estimated voxel-wise. This parameter transformation maintains the multi-compartment modeling capability while significantly reducing the optimization complexity and convergence time for high-resolution images.
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 high-resolution tissue parameterization with improved SNR and reduced acquisition time, enabling accurate characterization of tissue properties like white and gray matter, and cerebrospinal fluid densities, and fiber orientations.
Implementation Method 1
the presence of diffusion of water-molecules in the examined tissue is used to generate a contrast in magnetic resonances images
Implementation Method 2
diffusion-weighted magnetic resonance imaging, dMRI, which is an in-vivo and non-invasive imaging technique
Data Source
AI summary
Example embodiments describe a computer implemented method for obtaining parameterized characteristics of a tissue comprising obtaining at least two weighted MRI volume scans of the tissue, and transforming the two weighted MRI volume scans into parameters of a parameterized voxel-based model of the tissue by combined performing, by a super-resolution imaging technique, constructing a volume comprising voxels of the parameterized voxel-based model, and, by a quantitative MRI modelling technique, constructing the parameters for the respective voxels of the parameterized voxel-based model.