Disentangling True Brain Connectivity from Volume Conduction Artifacts

30 concepts30 papers1 proposal

When we record the brain's electrical or magnetic activity using Electroencephalography (EEG) or Magnetoencephalography (MEG), we are rarely looking at isolated signals. Instead, we observe a complex, blended tapestry. A major goal of modern cognitive neuroscience is to map functional connectivity—identifying how different brain regions communicate and synchronize with one another to process information.

However, a fundamental physical hurdle stands in the way: volume conduction (or spatial leakage). Because the head is a continuous, conductive medium composed of brain tissue, cerebrospinal fluid, skull, and scalp, electrical currents generated by a single neural source propagate widely through the volume of the head. Consequently, a single active brain region projects its signal to multiple sensors simultaneously.

This creates a dangerous illusion. If Sensor A and Sensor B both record the activity of Source X, their signals will appear highly synchronized. A naive analysis would conclude that the brain regions beneath Sensor A and Sensor B are actively communicating. In reality, this synchrony is a passive physical artifact, not a biological signal.

   [ Brain Source X ] 
       /        \
      /          \  (Instantaneous Volume Conduction)
     v            v
 [Sensor A]   [Sensor B]  <-- Artificially Correlated!

The Illusion of Synchrony

The impact of volume conduction spans across clinical and cognitive research. As reviewed by Esmaeil Seraj, volume conduction introduces spurious synchrony, leading researchers to identify functional connections that do not exist. This signal contamination distorts our understanding of healthy cognitive processes, such as the interhemispheric binding of visual motion studied by Gabriel N. Costa et al., and the tracking of perceptual closure loops detailed in Neuroelectromagnetic Correlates of Perceptual Closure Processes.

The problem is equally severe when trying to map pathological networks. For instance, studies on migraines by Chun‐Hsiang Chuang et al., stroke recovery by Sa Zhou et al., and childhood epilepsy by Hyun-Soo Choi et al. all note that passive signal leakage across the scalp obscures the true, direct pathways of neural communication. In psychiatric research, Golnoush Alamian et al. emphasize that magnetic field spread in MEG creates false correlations between adjacent brain regions, confounding the search for the electrophysiological signatures of depression and bipolar disorder.

To see how deceptive this can be, consider an invasive example: Laetitia Lalla et al. demonstrated that theta oscillations recorded in the rodent striatum—long assumed to represent local computations or inter-regional communication—were largely volume-conducted "mirages" leaking from the nearby hippocampus.

Exploiting Time Delays: Phase-Based Metrics

To separate true neural interaction from physical volume conduction, researchers exploit a key physical distinction: propagation speed.

Volume conduction is an electromagnetic field effect that propagates through head tissues at effectively the speed of light. For all practical purposes in electrophysiology, it is instantaneous, manifesting as a correlation at exactly zero time-lag.

In contrast, true biophysical communication across the brain is slow. It requires neurotransmitter release across synapses and action potential propagation along unmyelinated or myelinated axons. This introduces a physical time delay, which manifests in the frequency domain as a phase lag.

Volume Conduction: Source X ---> Sensor A & B (0 ms delay) -> Zero-Lag Correlation
True Interaction:  Source X ---> Source Y     (>0 ms delay) -> Phase-Lagged Correlation

By designing mathematical metrics that explicitly discard zero-lag correlations, researchers can filter out volume conduction.

Anti-Symmetric Cross-Correlation

Guido Nolte et al. introduced Interacting Source Analysis (ISA), a technique that isolates genuine neural communication by focusing exclusively on the anti-symmetric part of cross-correlation matrices. Because instantaneous mixtures from a common source are perfectly symmetric with respect to time reversal, any anti-symmetric component in the signal must stem from a time-delayed, genuine interaction.

Phase Lag Index (PLI)

Building on this intuition, metrics like the Phase Lag Index (PLI) quantify the asymmetry of the phase distribution between two signals. If two sensors are picking up a volume-conducted signal, their phase difference will cluster tightly around $0$ (or $\pi$). PLI discards these zero-lag components, measuring only the consistency with which one signal leads or lags another.

This approach has become vital for clinical biomarkers. For example, Lingyan Mao et al. and Masayoshi Oguri et al. successfully utilized PLI to bypass skull-blurring artifacts and uncover resting-state connectivity patterns that correlate with seizure frequency and severity in epilepsy patients. To further improve robustness against background noise, the Weighted Phase Lag Index (WPLI) scales the phase differences so that small, noise-induced phase rotations around zero have less impact. Chunying Fang et al. combined WPLI with topological data analysis to localize seizure onset zones with high accuracy.

Uncovering Causal Directionality

While phase-lag metrics tell us if an interaction is real, they often struggle to tell us which direction the information is flowing. However, traditional causal metrics like Granger Causality are highly sensitive to volume conduction.

As explained by Andreas Ziehe et al., when sensors capture an instantaneous mixture of underlying sources, standard Granger Causality suffers from high false-positive rates, often identifying a directed interaction where only common-source interference exists. To solve this, the authors compared Granger Causality to the Phase Slope Index (PSI). PSI calculates the slope of the phase difference as a function of frequency; a constant time delay creates a linear slope in the phase spectrum, allowing researchers to robustly infer directionality even in the presence of instantaneous mixtures.

Phase Difference
       ^
       |       /  <- Linear slope indicates true time delay (Directional Flow)
       |      /
       |----/-----> Frequency
       |

Other information-theoretic approaches bypass linear assumptions entirely. Raúl Vicente et al. demonstrated that Transfer Entropy (TE) can successfully map directed information flow in MEG recordings. Because Transfer Entropy measures how much the past of Signal A helps predict the future of Signal B (beyond what the past of Signal B can predict on its own), it remains inherently robust against the zero-lag, instantaneous correlations caused by volume conduction. Similarly, Eric Rawls et al. developed a computational framework to isolate rhythmic oscillations from background noise, allowing the identification of distinct causal pathways in traumatic brain injury patients that were previously invisible due to signal bleeding.

Source-Space Leakage Correction

An alternative strategy is to move from "sensor space" (the electrodes on the scalp) to "source space" (estimating the actual dipoles inside the brain) using mathematical inversion techniques like beamforming or Minimum Norm Estimation (MNE), as explored by Wenjie Li et al.

However, source localization does not completely eliminate the problem. Because the inverse problem is mathematically underdetermined, the reconstructed sources still suffer from spatial leakage—the activity of a single source bleeds into neighboring reconstructed voxels.

Mark W. Woolrich et al. evaluated the reliability of MEG resting-state metrics and discovered that many popular methods exhibit artificially high consistency precisely because of this residual spatial leakage. To combat this, they recommend using leakage-corrected metrics, such as partial correlation or amplitude envelope correlation combined with multivariate orthogonalization. This mathematical step strips away any shared, instantaneous signal between two source locations before computing their connectivity.

Furthermore, traditional spatial filters often fail when true neural sources are highly synchronized, as the algorithm can mistake the correlated biological signals for an external noise source and suppress them. To address this limitation, Johanna M. Zumer et al. developed a probabilistic graphical model that treats noise and interference as hidden variables learned from pre-stimulus data. This allows the algorithm to simultaneously localize brain activity and suppress volume conduction noise, even when multiple, temporally correlated brain sources are firing in sync.

By combining physical intuition about time delays with sophisticated source-space orthogonalization, modern electrophysiology can look past the physical mirage of volume conduction to map the true, dynamic networks of the human brain.

Key Foundational Work

Analyzing Coupled Brain Sources: Distinguishing True from Spurious Interaction (2005) — contributed 1 concept elements to this direction.
A Probabilistic Algorithm Integrating Source Localization and Noise Suppression of MEG and EEG data (2006) — contributed 1 concept elements to this direction.
History & citationVersion 2 · 2026-10-04

Current · Permanent link

  1. Version 2 (latest) · 2026-10-04 · 26 papers
  2. Version 1 · 2026-10-04 · 25 papers
Cite this direction

Latest version (a living page):

@misc{lacuna_volume_conduction_artifact_23d8ad3f,
  title = {volume conduction artifact},
  author = {{Lacuna}},
  howpublished = {Lacuna research direction},
  year = {2026},
  url = {https://lacuna.tiptreesystems.com/direction/volume-conduction-artifact/rdr_23d8ad3f8ac54e91b3e1bd937637810d},
  note = {Latest version of a living research direction; cite a version for a fixed text. Accessed 2026-11-19},
}

This exact version (version 2, frozen text):

@misc{lacuna_volume_conduction_artifact_23d8ad3f_v2,
  title = {volume conduction artifact},
  author = {{Lacuna}},
  howpublished = {Lacuna research direction},
  year = {2026},
  url = {https://lacuna.tiptreesystems.com/direction/volume-conduction-artifact/rdr_23d8ad3f8ac54e91b3e1bd937637810d/v/2},
  note = {Version 2, frozen 2026-10-04. Accessed 2026-11-19},
}
About

These directions overlap with this one through shared papers.

See something wrong? Report a mistake