Dynamic functional graph-Laplacian priors integrated with optimization for EEG source localization

Clinical Snapshot

Key Findings

• P Value: Not reported in abstract

• Effect Size: Binary left-versus-right MI classification accuracy of 93.52% on PhysioNet dataset, reported as superior to representative deep learning baselines; magnitude of improvement over comparators not quantified in abstract

• Primary Outcome: Centre-of-mass localization error in synthetic benchmarks: DG-ADMM achieved lowest mean error in 5 of 8 conditions (single-source, double-source, correlated-source, and dynamic-source at two SNR levels)

• Nnt Or Sensitivity: Not applicable in standard clinical sense; localization error metric: lowest in 5/8 synthetic conditions; classification accuracy: 93.52% (no sensitivity/specificity breakdown provided)

• Confidence Interval: Not reported

Clinical Application

Computational feasibility is claimed but not benchmarked with runtime data. The pipeline requires LCMV beamforming, PCA, sliding-window PLV with surrogate testing, and ADMM optimisation — a multi-stage process that may not be compatible with real-time BCI requirements without significant engineering optimisation. Implementation requires EEG systems with sufficient channel density, accurate head models (MRI co-registration), and expertise in signal processing. EEG-based BCI technology is not currently listed on the Australian PBS or approved by the TGA as a standalone therapeutic device. In the Australian context, EEG source localization is used clinically in epilepsy pre-surgical evaluation (available at major neuroscience centres including Austin Health, Royal Melbourne Hospital, and Westmead Hospital) and in research settings for rehabilitation BCI. RACGP guidelines do not currently address BCI-assisted motor rehabilitation. The PhysioNet dataset used is internationally derived; Australian-specific validation would be required before clinical adoption. The National Disability Insurance Scheme (NDIS) may be a relevant funding pathway for BCI-assisted rehabilitation if clinical efficacy is established through prospective trials. TGA regulatory classification of BCI software as a Software as a Medical Device (SaMD) would apply under the current regulatory framework. Individuals undergoing EEG-based brain-computer interface assessment, particularly for motor imagery decoding in the context of motor rehabilitation (e.g., post-stroke, spinal cord injury). Potentially applicable to epilepsy source localization and pre-surgical mapping, though not evaluated in these contexts.

Abstract

Objective. Electroencephalography (EEG) source localization is an ill-posed inverse problem in which conventional methods often rely on static anatomical or smoothness assumptions and may neglect task-related dynamic functional interactions. This study aims to develop a dynamic graph-regularized EEG source localization framework that incorporates time-varying functional connectivity directly into inverse reconstruction and improves source-space motor imagery (MI) decoding.Approach. We propose DynaGraph-alternating direction method of multipliers (DG-ADMM), a source localization framework that combines linearly constrained minimum variance (LCMV) beamforming, region-of-interest-level dimensionality reduction, dynamic phase synchronization analysis, graph-Laplacian regularization, and efficient optimization. Initial source estimates are obtained using a LCMV beamformer. Region-level source signals are then extracted using principal component analysis and sliding-window phase-locking values with surrogate-based statistical testing are used to construct reliable dynamic functional graphs. The resulting graph Laplacian is mapped back to source space and embedded as a structured prior in an ADMMs optimization problem.Main results. Experiments on the MNE sample dataset showed that DG-ADMM produced spatially concentrated and physiologically plausible source patterns. On the PhysioNet MI dataset, the proposed framework achieved a binary left-versus-right MI classification accuracy of 93.52%, outperforming representative deep learning baselines. In a 320-dataset synthetic benchmark covering single-source, double-source, correlated-source, and dynamic-source conditions at two signal-to-noise ratios, DG-ADMM achieved the lowest mean center-of-mass localization error in five of eight conditions and showed its clearest advantage for dynamic sources.Significance. The results demonstrate that dynamic functional connectivity can serve as an informative graph-structured prior for EEG inverse reconstruction. DG-ADMM provides an interpretable and computationally feasible strategy for improving spatial focus, temporal consistency, and source-space decoding performance in EEG-based brain-computer interfaces.

References

1. 1.Fei, S.-W., Chen, Y., & Chen, J.-L. (2026). Dynamic functional graph-Laplacian priors integrated with optimization for EEG source localization. Journal of Neural Engineering. https://doi.org/10.1088/1741-2552/ae857e