FINES Algorithm for Neural Source Localization
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current source localization methods, such as MUSIC, face challenges in accurately determining neural activity locations in the brain due to biased estimates when sources are weak or highly correlated, and are affected by spatially correlated brain noise, which reduces spatial resolution and accuracy.
Innovation Solution
The FINES algorithm employs a novel subspace source localization approach by projecting vectors onto a subspace spanned by a set of particular vectors in the estimated noise-only subspace, which is closest to the array manifold associated with a brain region, enhancing the accuracy and spatial resolution of dipole source localization.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If MUSIC algorithm is used for source localization, then source locations can be determined from EEG or MEG signals, but spatial resolution and localization accuracy deteriorate in the presence of spatially correlated brain noise
Solution Approach 1:
The FINES algorithm extracts and removes spatially correlated brain noise from the recorded signals by estimating the noise subspace and projecting the data onto this subspace to eliminate noise components before performing source localization, thereby improving localization accuracy in noisy conditions
Solution Approach 2:
The algorithm changes the parameter representation by using a subspace projection approach with optimized projection vectors that are specifically designed to be orthogonal to the noise subspace, transforming the source localization problem into a noise-robust estimation problem that maintains accuracy despite the presence of spatially correlated noise
2Measurement precision
If traditional subspace methods like MUSIC are used, then source locations can be estimated, but estimation bias increases when sources are weak or highly correlated
Solution Approach 1:
The FINES algorithm performs preliminary noise subspace estimation and projection before source localization, preparing a noise-orthogonal basis that eliminates correlated noise effects in advance, which prevents estimation bias from developing during the source localization process itself
Solution Approach 2:
The algorithm introduces an intermediary noise-orthogonal projection step between signal acquisition and source localization, using projected data that has been transformed to remove noise correlations, thereby mediating the effect of noise on source estimation and reducing bias
3Object-affected harmful factors
If noise covariance matrix incorporation is used to deal with spatially correlated noise, then noise robustness may improve, but the required noise covariance matrix is unknown or difficult to estimate in most experimental conditions
Solution Approach 1:
The FINES algorithm makes the system self-sufficient by estimating the noise subspace directly from the recorded data without requiring external noise covariance information, allowing the algorithm to adapt to different experimental conditions automatically without complex pre-characterization of noise properties
Data Source
AI summary
Described herein is a non-invasive determination of locations of neural activity in a brain. In particular, methods and systems have been developed that utilize a FINES algorithm for use in three-dimensional (3-D) dipole source localization to locate neural activity in a brain.