Adaptive Brain Voxelization for EEG Simulation Accuracy

Solution Overview

Problem

Existing methods for partitioning a human brain volume in medical images, such as those used for simulating electrical brain activity, face challenges including incomplete spatial representation, high computational cost, and inaccurate localization of epileptogenic zones due to anatomically-based parcellations, which are not optimized for electrode placements.

Innovation Solution

A method for voxelizing a 3D structural medical image of the brain, where each voxel is subdivided based on electrode sensor locations, ensuring accurate representation and reduced computational overhead by optimizing voxel size and coverage relative to electrode detection ranges, thereby improving the simulation of electrical brain activity and localization of epileptogenic zones.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Ease of operation

If anatomically-based parcellation methods are used, then the partitioning results are easy to understand by clinicians, but the simulation performance varies and computational cost increases

Engineering Contradiction:

Improveease of understanding partitioning resultsVSAvoidsimulation performance

Core Design Contradiction:

Ease of operationVSProductivity

Solution Approach 1:

The patent applies local quality by creating a non-uniform voxel grid where voxel size varies depending on its spatial location relative to electrode sensors. Voxels near electrodes are smaller to capture local electrical activity details, while voxels farther away are larger to reduce overall computational load. This spatially-dependent voxel sizing optimizes simulation performance for each region based on its functional importance for electrode signal generation.

Inventive Principle:

Principle #3Local quality

Solution Approach 2:

The patent segments the brain volume into multiple voxels of varying sizes based on their distance from electrode sensors. Instead of using a uniform anatomical parcellation, the brain is divided into fine-grained voxels near electrodes and coarser voxels farther away, creating a hierarchical segmentation that balances computational efficiency with simulation accuracy for the specific electrode configuration.

Inventive Principle:

Principle #1Segmentation

2Device complexity

If large anatomical regions are used in parcellation, then the number of regions is reduced, but the inverse problem in EEG source analysis becomes over-constrained

Engineering Contradiction:

Improvenumber of regionsVSAvoidinverse problem solvability

Core Design Contradiction:

Device complexityVSReliability

Solution Approach 1:

The patent applies local quality by creating a non-uniform voxel grid where voxel size varies depending on its spatial location relative to electrode sensors. Voxels near electrodes are smaller to capture local electrical activity details, while voxels farther away are larger to reduce overall computational load. This spatially-dependent voxel sizing optimizes simulation performance for each region based on its functional importance for electrode signal generation.

Inventive Principle:

Principle #3Local quality

Solution Approach 2:

The patent dynamically adjusts voxel sizes based on the specific electrode sensor configuration and the spatial relationship between voxels and electrodes. The voxelization scheme adapts to the particular imaging and electrode setup, creating an optimal parcellation for each case rather than using a fixed anatomical atlas, thereby preventing over-constraint in the inverse problem.

Inventive Principle:

Principle #15Dynamics

3Area of stationary object

If anatomical regions far from electrode sensors are included, then the simulation covers the entire brain, but computational cost increases due to signal attenuation

Engineering Contradiction:

Improvebrain coverageVSAvoidcomputational cost

Core Design Contradiction:

Area of stationary objectVSUse of energy by moving object

Solution Approach 1:

The patent segments the brain volume into multiple voxels of varying sizes based on their distance from electrode sensors. Instead of using a uniform anatomical parcellation, the brain is divided into fine-grained voxels near electrodes and coarser voxels farther away, creating a hierarchical segmentation that balances computational efficiency with simulation accuracy for the specific electrode configuration.

Inventive Principle:

Principle #1Segmentation

Solution Approach 2:

The patent changes the parameter of voxel size based on spatial location and distance from electrode sensors. By dynamically adjusting voxel dimensions, the simulation can maintain high resolution near electrodes where signal generation is critical, while using coarser voxels in distant regions where signal attenuation makes those areas less relevant, thereby reducing overall computational cost.

Inventive Principle:

Principle #35Parameter changes

Data Source

PatentEP4120201B1Voxelization of a 3D structural medical image of a human brain

Publication Date: 2026.02.11 DASSAULT SYSTEMES SA

Figure 1
Figure 2
Figure 3

AI summary

The invention notably relates to a computer-implemented method for voxelizing a 3D structural medical image of a human's brain. The method comprises providing a 3D structural medical image of the human's brain, comprising a reference frame; generating a voxelized 3D structural medical image; providing parameters of at least one EEG electrode sensor and, for each EEG electrode sensor: a localization in the voxelized 3D structural medical image's reference frame; and a sensor detection distance; providing a regular 3D grid of voxels; and for each voxel of the 3D grid, iteratively subdividing the voxel while the distance between the voxel and the localization of any electrode sensor is smaller or equal than the sensor detection distance and while a size of the voxel is greater than a predetermined length, each subdivided voxel joining a finite number of voxel(s) of the voxelized 3D structural medical image.