Sensitivity of EEG leads to volume conductor properties

Sven Wagner, Johannes Vorwerk, Lars Ruthotto, Harald Kugel, M. Burger, Thomas R. Knösche, Burkhard Maeß, Carsten H. WoltersBiomedizinische 2012 · DOI

Abstract

Accurate EEG source localization depends heavily on the forward model used to describe how electrical signals propagate from the brain to the scalp. This study investigates the sensitivity of EEG lead fields to the anatomical complexity of the head model. By progressively adding realistic tissue layers—including the cerebrospinal fluid (CSF), distinct skull layers, and white matter anisotropy—to a Finite Element Method (FEM) model, we demonstrate that simplified three-layer models significantly misrepresent the volume conductor properties. Our results show orientation shifts of over 50 degrees and magnitude changes of up to 250%, suggesting that high anatomical detail is essential for reliable brain mapping.

Introduction

To solve the "inverse problem" in EEG—identifying the neural origins of recorded scalp potentials—researchers must first solve the "forward problem." This requires a mathematical description of the head as a volume conductor. Historically, computational constraints led to the widespread adoption of simplified three-compartment models consisting of the skin, skull, and brain. However, these models ignore the complex internal architecture of the human head, such as the high conductivity of the cerebrospinal fluid (CSF) and the multi-layered nature of the skull. This paper evaluates how these simplifications affect the resulting lead fields and, by extension, the accuracy of source reconstruction.

Methodology

The researchers employed an "effect-by-effect" analysis using high-resolution Finite Element Method (FEM) head models. Starting from a baseline isotropic three-compartment model, complexity was added incrementally:

  1. Six-Compartment Expansion: The model was refined to include the skin, skull compacta (hard bone), skull spongiosa (spongy bone), CSF, grey matter, and white matter.
  2. Conductivity Anisotropy: The model incorporated brain conductivity anisotropy, accounting for the fact that electrical current travels more efficiently along the longitudinal axis of white matter fibers than across them.
  3. Lead Field Analysis: The study assessed changes in the lead vector field at each stage of complexity. The primary metrics for comparison were the change in vector orientation (angle) and the change in vector magnitude (strength) within the cortical volume.
  4. Visualization: Computed vector fields were visualized directly to observe how different tissues influence the path of least resistance for electrical currents.

Key Insights: The "Highway" Effect

The study identified a significant "channeling effect" caused by low-resistive tissues. Regions such as the skin, skull spongiosa, and particularly the CSF act as electrical "highways."

  • Path of Least Resistance: In areas of high resistance, such as the skull compacta, electrical vectors tend to bend toward the nearest low-resistance compartment.
  • CSF Influence: The highly conductive CSF redistributes the lead field significantly, a factor often overlooked in models that treat the entire intracranial space as a single homogeneous compartment.
  • Anisotropy: When white matter anisotropy is included, the electrical vector field aligns more closely with the physical orientation of nerve fibers (conductivity tensors), further refining the accuracy of the lead field.

Results

The transition from a simplified three-layer model to a realistic six-compartment anisotropic model yielded dramatic differences in the forward solution:

  • Orientation Shifts: The orientation of lead vectors in the cortical compartment shifted by more than 50 degrees. Such a large angular error suggests that simplified models may misattribute the direction of neural dipoles.
  • Magnitude Fluctuations: The magnitude of the lead vectors changed by up to 250%. This indicates that simplified models can severely over- or under-estimate the strength of neural sources.
  • Tissue-Specific Impact: The inclusion of the skull spongiosa and CSF were identified as the most critical factors in reducing the error of the forward model, as these layers provide the primary pathways for current channeling.

Discussion and Conclusion

The findings of this study suggest that the use of simplified three-layer head models is not merely a minor approximation but a fundamental mischaracterization of the signal's path from the brain to the scalp. The massive shifts in orientation and magnitude demonstrate that anatomical complexity is mandatory for high-precision EEG source localization.

While the study highlights the necessity of these complex models, it also points toward the need for advanced visualization tools. By observing lead vector fields directly, researchers can better understand the volume conduction effects within their specific head models. Future work should focus on quantifying the direct impact of these forward modeling errors on the localization error (in millimeters) during inverse reconstruction and optimizing the computational efficiency of six-compartment anisotropic FEM solvers.

Summary updated

Abstract

1 Institute for Biomagnetism and Biosignalanalysis, University of Muenster, 48149, Muenster, Germany 2 Institute of Mathematics and Image Computing, University of Luebeck, 23562 Luebeck, Germany 3 Department of Clinical Radiology, University of Muenster, 48149, Muenster, Germany 4 Institute for Computational and Applied Mathematics, University of Muenster, 48149, Muenster, Germany 5 Max Planck Institute for Human Cognition and Brain Sciences, PO box 500355, 04303 Leipzig, Germany
BibTeX
@inproceedings{wagner2012sensitivity,
  author = {Sven Wagner and Johannes Vorwerk and Lars Ruthotto and Harald Kugel and M. Burger and Thomas R. Knösche and Burkhard Maeß and Carsten H. Wolters},
  title = {{Sensitivity of EEG leads to volume conductor properties}},
  booktitle = {Biomedizinische},
  year = {2012},
  doi = {10.1515/bmt-2012-4065},
}
Figure