Rotational-Invariant Diffusion Fitting With Precomputed Signal Dictionaries
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing diffusion imaging methods struggle to reliably and efficiently estimate muscle fiber diameter using a rotationally-invariant method, leading to computational inefficiencies and poor convergence in fitting models.
Innovation Solution
A computer-implemented method involving the generation of a set of simulated diffusion signals and metrics, consolidation into an orientationally-invariant dictionary, and application of this dictionary to acquired diffusion MR signals to generate macroscopic diffusion features, such as apparent fiber diameter and fluid fraction, using techniques like DTI fitting and model-based de-noising.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If conventional diffusion imaging fitting methods are used, then model fitting can be performed, but computational efficiency is poor and convergence is unreliable
Solution Approach 1:
The patent pre-computes and stores a dictionary of diffusion signals and corresponding microstructural parameters before actual data analysis. This preliminary action allows the fitting process to simply lookup pre-computed values rather than performing complex iterative optimizations, dramatically improving both reliability and speed.
Solution Approach 2:
The patent creates a comprehensive dictionary that copies and stores pre-computed diffusion signals for various microstructural parameters. Instead of重新computing these signals during fitting, the system copies and matches acquired signals against this pre-existing dictionary, eliminating redundant computations and ensuring consistent results.
2Adaptability or versatility
If rotationally-invariant methods are used to estimate muscle fiber diameter, then orientation independence is achieved, but computational complexity increases
Solution Approach 1:
The patent merges diffusion signals from multiple orientations by computing signal averages across different gradient directions. This combining approach creates rotationally-invariant metrics that represent tissue properties independent of fiber orientation, while the pre-computed dictionary handles the complexity of multi-orientation processing.
Solution Approach 2:
The patent transforms the fitting problem from estimating orientation-dependent parameters to estimating orientation-invariant parameters. By changing the parameter space to use rotationally-invariant metrics (such as mean diffusivity and fractional anisotropy), the system achieves orientation independence without requiring complex orientation-specific computations.
Data Source
AI summary
A computer-implemented method includes: generating a set of simulated diffusion signals based on a corresponding set of diffusion sampling parameters and a corresponding set of micro structural model parameters, wherein the diffusion sample parameters correspond to magnetic resonance (MR) parameters, and wherein the microstructural model parameters characterizing a microscopic diffusion with a spatial orientation; processing the set of simulated diffusion signals based on the corresponding set of diffusion sampling parameters to generate a first set of output metrics, wherein the first set of output metrics are associated with the spatial orientation; consolidating multiple sets of output metrics into a dictionary, wherein each set of the multiple sets of output metrics are generated by processing a corresponding set of simulated diffusion signals and associated with a corresponding spatial orientation; applying the dictionary to a set of acquired diffusion MR signals; and generating at least one macroscopic diffusion feature that is orientation invariant.