Multi-Compartment Diffusion Model Averaging for MRI Signal Analysis

Resolve Bottlenecks,
Find Innovative Solutions
Generate Solutions

Solution Overview

Problem

Current methods for characterizing molecular diffusion in the body using diffusion-weighted magnetic resonance signals face limitations in spatial resolution, leading to challenges in determining white matter microstructure and tractography due to hardware constraints and computational inefficiencies in model selection and estimation.

Innovation Solution

The method employs model averaging of multi-compartment diffusion models with varying numbers of compartments, converting them into extended models with a fixed number of compartments for averaging, and simplifies the result using clustering to determine the optimal number of compartments, thereby improving the characterization of molecular diffusion profiles.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Measurement precision

If brute-force methods are used to solve the model selection problem by fitting nested candidate MCMs with increasing number of fascicles, then the model selection can be performed, but the method tends to favor MCMs that over-fit the signals and requires extensive computational resources

Engineering Contradiction:
Improvemodel selection accuracyVSAvoidcomputational complexity
Core Design Contradiction:
Measurement precisionVSDevice complexity

Solution Approach 1:

The patent applies preliminary action by performing model selection through a two-stage process: first selecting from a reduced candidate set based on preliminary criteria, then refining the selection. This avoids the computational burden of evaluating all possible nested MCM configurations while maintaining selection accuracy, thereby resolving the contradiction between measurement precision and device complexity.

Inventive Principle:
Principle #10Preliminary action

Solution Approach 2:

The patent segments the model selection process into distinct stages: initial model candidate generation, preliminary filtering based on fit criteria, and final selection. This segmentation reduces the computational complexity by breaking down the overwhelming task of evaluating all nested MCMs into manageable steps, while still achieving accurate model selection.

Inventive Principle:
Principle #1Segmentation

2Measurement precision

If the number of compartments in MCM is increased to better fit the diffusion signals, then the model fit improves, but the model complexity increases leading to overfitting

Engineering Contradiction:
Improvesignal fit accuracyVSAvoidmodel generalization
Core Design Contradiction:
Measurement precisionVSReliability

Solution Approach 1:

The patent applies partial action by considering a predefined set of candidate MCMs with a limited range of compartment numbers rather than evaluating all possible configurations. This partial exploration of the model space is sufficient to identify the optimal model while avoiding the excessive complexity that leads to overfitting, thus balancing signal fit accuracy with model generalization.

Inventive Principle:
Principle #16Partial or excessive action

Solution Approach 2:

The patent changes the parameter of model complexity by systematically varying the number of compartments across candidate MCMs and selecting the optimal value based on fit criteria. This controlled parameter change allows the model to achieve sufficient signal fit accuracy without increasing complexity to the point of overfitting, thereby maintaining model reliability.

Inventive Principle:
Principle #35Parameter changes

3Extent of automation

If Bayesian frameworks are used to estimate the best MCM by maximizing posterior distribution, then model selection and estimation are performed simultaneously, but the methods are prohibitively computationally expensive

Engineering Contradiction:
Improveautomated model selectionVSAvoidcomputational cost
Core Design Contradiction:
Extent of automationVSDevice complexity

Solution Approach 1:

The patent extracts the computationally intensive posterior maximization step from the automated model selection process and replaces it with more efficient criteria-based selection. This extraction maintains the automation benefit while removing the prohibitive computational cost, resolving the contradiction between extent of automation and device complexity.

Inventive Principle:
Principle #2Taking out (Extraction)

Solution Approach 2:

The patent uses computationally inexpensive model selection criteria instead of expensive Bayesian posterior maximization. These simpler, more efficient methods achieve sufficient model selection automation without the prohibitive computational cost, effectively replacing complex Bayesian frameworks with lighter alternatives.

Inventive Principle:
Principle #27Cheap short-living objects (Disposable)

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 enhances the characterization of molecular diffusion by making better use of available information, reducing computational complexity, and providing a more robust and efficient method for determining white matter microstructure and tractography compared to existing brute-force and Bayesian methods.

Implementation Method 1

the application of a magnetic field spatial gradient, hereafter referred to as a diffusion-sensitizing gradient (DSG), wherein the intensity of each voxel is proportional to how far water molecules in this voxel moved along the DSG direction

Methodology Applied
Scientific EffectMagnetic field gradient: Magnetic Field

Data Source

PatentEP2933743B1Method of characterizing molecular diffusion within a body from a set of diffusion-weighted magnetic resonance signals and apparatus for carrying out such a method
Publication Date: 2019.05.08 CENT NAT DE LA RECH SCI (C N R S)
  • EP2933743B1 patent drawingFigure 1~5
  • EP2933743B1 patent drawingFigure 3
  • EP2933743B1 patent drawingFigure 4a~4b

AI summary

A computer-implemented method of characterizing molecular diffusion within a body from a set of diffusion-weighted magnetic resonance signals by computing a weighted average (AVM) of a plurality of multi-compartment diffusion models (EXM1, EXM2) comprising a same number of compartments, fitted to a set of diffusion-weighted magnetic resonance signals, said weighted average being computed using weights representative of a performance criterion of each of said models; wherein each of said multi-compartment diffusion models comprises a different number of subsets of compartments, the compartments of a same subset being identical to each other.