EEG Source Separation Using Physically Unique Artifact Dictionaries
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current methods for separating cerebral sources from artifactual sources in EEG recordings are not completely effective, particularly during procedures like EEG conducted during MRI, where artifacts from pulsating blood flows and electrical noise interfere with neuronal signal analysis.
Innovation Solution
A method using physically unique dictionary elements for signal separation, where a coupled lead forward matrix is formed with 'good' and 'bad' atoms to represent intracranial and extracranial sources, allowing for sparse coding to isolate and remove artifacts from the signal, ensuring the dictionary is overcomplete to maximize sparseness of coefficients.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If statistical decomposition or independent component analysis is used to remove artifacts from EEG data, then some artifact components can be identified and subtracted, but the separation is not completely effective and artifactual contamination remains
Solution Approach 1:
The patent segments the signal sources into two distinct categories: intracranial sources (neurons inside the skull) and extracranial sources (artifacts outside the skull). By creating separate template sets for each category and solving for their respective coefficients independently, the method achieves complete separation of artifact components from neuronal signals, resolving the incomplete separation problem of prior art.
Solution Approach 2:
The patent applies local quality by creating physically unique templates that are specifically tuned to the distinct physical properties of intracranial versus extracranial sources. The templates encode different spatial frequency characteristics and propagation paths, allowing the method to selectively identify and remove only extracranial artifacts while preserving intracranial neuronal activity with high fidelity.
2Object-affected harmful factors
If standard artifact removal techniques are applied to EEG recorded during MRI, then some ballistocardiographic artifacts can be reduced, but the strong magnetic field environment creates additional artifacts that contaminate the signal
Solution Approach 1:
The patent converts the harmful ballistocardiographic artifacts into beneficial information by creating extracranial templates that specifically model the spatial frequency characteristics of pulsating blood flow artifacts in the MRI environment. By solving for extracranial coefficients and nullifying these components, the method transforms the previously harmful strong magnetic field artifacts into identifiable patterns that can be systematically removed, thereby improving neuronal signal detection accuracy.
3Productivity
If the inverse problem is solved using measured potentials that contain artifacts, then source locations can be estimated, but the accuracy is compromised because the measured potentials have multiple causes
Solution Approach 1:
The patent performs preliminary action by completely separating and removing extracranial artifact components from the measured EEG potentials before solving the inverse problem. By nullifying extracranial coefficients in the template matrix and computing only intracranial source contributions, the method ensures that the inverse problem is solved using cleaned, artifact-free potentials, thereby maximizing source location accuracy without sacrificing computational efficiency.
Data Source
AI summary
A method for separating signal sources by use of physically unique dictionary elements. The method is particularly advantageous for separating cerebral from artifactual sources in electroencephalographic (EEG) recording, making use of dictionary element pattern recognition methods that are tuned to the unique physical properties of each source domain.