Improving EEG brain source mapping with refined and completed signal components
Observability-Aligned Diffusion and Null-Space Completion for EEG Source Localization
Computational Engineering, Finance, and Science
Summary
Identifying where brain signals come from using EEG is tricky because the measurements don’t capture all the details needed. The authors propose a method that first sharpens the parts of the signal that can be observed, then intelligently fills in the missing parts based on learned patterns. They train their system using simulated brain activity without needing real paired EEG and brain data. Tests show their method finds brain activity locations more accurately than standard approaches, and initial trials with real brain signals look promising.
What this means in practice
- •For clinical neurophysiology teams: Improve the spatial accuracy of noninvasive brain signal localization from scalp EEG data during neurological assessments.
- •For brain-computer interface developers: Enhance decoding of user intentions by better mapping EEG sources using a combined learned and deterministic signal refinement approach.
Authors
Zirui Wen, Shihao Yang, Feng Liu
Abstract
Electrophysiological source imaging is severely underdetermined because the lead-field matrix has limited rank relative to the source dimension and strongly correlated columns. Scalp EEG measurements constrain only the observable row-space component, leaving a high-dimensional null space unobservable. We propose a two-stage framework coupling posterior diffusion refinement with deterministic null-space completion. A minimum-norm-initialized, singular-coordinate Posterior-DDIM sampler combines a spatiotemporal diffusion prior with direction-dependent measurement reliability to refine the observable component. A frozen ridge predictor and nonlinear temporal residual network then estimate the null component under explicit projection constraints, preserving the refined row estimate and its predicted EEG exactly. All learned components are trained exclusively on simulated cortical activity and its row-null decomposition, without paired EEG-source training data. Across simulations with varying noise levels, the method achieved higher frame-level area under the receiver operating characteristic curve and average precision, together with lower localization error, than conventional and learning-based baselines. Exploratory evaluations on evoked EEG and simultaneous scalp EEG-stereo-EEG recordings showed condition- and event-dependent concordance with cross-modal and intracranial references.