Coupled CP Decomposition of Simultaneous MEG-EEG Signals for Differentiating Oscillators During Photic Driving

Naskovska, Kristina; Lau, Stephan; Korobkov, Alexey A.; Haueisen, Jens; Haardt, Martin · 2020 · Crossref

DOI: 10.3389/fnins.2020.00261

archive: archived pipeline: cataloged verified

Get this paper ↗ (DOI — opens at the source; we link to it, we don't host it)

Summary

This study addresses the challenge of extracting and differentiating signal sources from simultaneous magnetoencephalography (MEG) and electroencephalography (EEG) data during intermittent photic stimulation (IPS). Because MEG and EEG capture different physical aspects of the same neuronal activity, the resulting data are inherently multi-dimensional and coupled. The authors aim to leverage this coupling to isolate physiologically meaningful oscillators, specifically investigating frequency entrainment and resonance effects in the alpha and theta bands. The motivation stems from the need for robust, unsupervised methods to analyze complex, multi-modal biomedical signals for clinical diagnostics and brain-computer interfaces. The researchers employed the Coupled Semi-Algebraic framework for approximate CP decomposition via SImultaneous matrix diagonalization (C-SECSI). This method jointly decomposes heterogeneous tensors that share at least one factor matrix, offering advantages in ill-conditioned scenarios with collinear factors or varying noise varances. The study first validated C-SECSI against alternative methods using simulated benchmark data. Subsequently, the method was applied to simultaneous MEG-EEG recordings from 12 healthy participants. Participants underwent IPS with stimulation frequencies ranging from 0.4 to 1.3 times their individual alpha frequency. Signals were processed using complex Morlet wavelet decomposition to estimate instantaneous frequencies, and the resulting tensors were normalized to align amplitude scales before decomposition. The analysis assumed a tensor rank of two, identifying components based on reconstruction error and reliability metrics. The benchmark tests demonstrated that C-SECSI is more accurate than standard SECSI and other alternative methods, particularly in scenarios involving noise sources with different variances or highly correlated factors. In the experimental analysis, the component field-maps successfully separated visually evoked brain activity from background signals. The frequency signatures of the extracted components identified either entrainment to the stimulation frequency or its first harmonic, or oscillations within the individual alpha or theta bands. Crucially, the group analysis revealed a reciprocal relationship between alpha and theta band oscillations across both MEG and EEG data. The study concludes that coupled tensor decomposition using C-SECSI is a robust and powerful tool for extracting physiologically meaningful sources from multidimensional biomedical data. By effectively handling the coupling between MEG and EEG, the method provides a reliable solution for unsupervised signal source extraction. This capability is significant for advancing the accessibility of multi-modal signal acquisition technologies in clinical diagnostics, pre-surgical planning, and brain-computer interface applications, where distinguishing specific neural oscillators from noise is critical.

Provenance

The full processing record for this entry. Every stage of this paper's journey through the pipeline is logged — what ran, with which tool and model, how many attempts it took, and when it last completed.

Summary generated by qwen3.6-27b-nvidia on 2026-08-10; verification: verified.

Topics

Ranked by relevance to this paper. Hover a topic for its definition.

• neuro workload indices