Electroencephalography

Electroencephalography (EEG) is a non-invasive electrophysiological recording modality that measures fluctuating microvolt-scale electric potential differences across the scalp generated by synchronous transmembrane ionic currents in millions of cortical pyramidal cells. First recorded from the intact human head on July 6, 1924 by German psychiatrist Hans Berger at the University of Jena—using silver-foil electrodes coupled to a Siemens double-coil galvanometer—and published in Berger's landmark 1929 report Über das Elektrenkephalogramm des Menschen, the method demonstrated that the waking human brain generates continuous rhythmic oscillations shifting between relaxed eyes-closed wakefulness (alpha waves, 8–13 Hz) and active mental calculation (beta waves, 13–30 Hz). Validated by Edgar Douglas Adrian and Bryan H. C. Matthews in 1934, scalp EEG delivers millisecond temporal resolution without ionizing radiation or surgery across clinical epileptology, sleep staging, intraoperative monitoring, cognitive neuroscience, and non-invasive brain–computer interfacing.

Apical Dendritic Postsynaptic Currents in Cortical Pyramidal Layers

Although cortical neurons transmit long-range outputs via all-or-none spikes, scalp sensors rarely register isolated Action Potential waveforms. Somatic sodium–potassium spikes are brief (~0.5–1.5 ms), reverse polarity too rapidly to sum without submillisecond alignment, and create compact quadrupolar extracellular fields that decay steeply with distance (1/r³). Conversely, excitatory and inhibitory postsynaptic potentials evoked during Synaptic Transmission last 10 to 150 ms, enabling thousands of inputs across neighboring columns to sum in phase.

The principal cellular generators of scalp EEG are layer III and layer V cortical pyramidal neurons, whose elongated apical dendrites extend perpendicular to the pial surface in parallel palisades. Excitatory glutamatergic activation of AMPA and NMDA receptors on superficial layer I apical tufts drives Na⁺ and Ca²⁺ influx, forming an extracellular current sink (negative potential) near the pia balanced by passive capacitive and ohmic return sources exiting the deeper soma and basal dendrites. Perisomatic GABA_A Cl⁻ influx creates an identically oriented dipole of opposite sign, preventing scalp polarity alone from distinguishing superficial excitation from deep inhibition. Because radially symmetric stellate cells form closed-field geometries whose external currents cancel, measurable scalp voltages (10 to 100 μV) require synchronous dipolar activity across 6 to 10 cm² of cortex (~10⁷ to 10⁸ neurons; Buzsáki et al., 2012).

Poisson Forward Modeling and Cranial Spatial Blurring

Below ~1 kHz, capacitive and inductive propagation effects in head tissues are negligible, reducing Maxwell's equations to the quasi-static Poisson volume-conduction equation:

∇ · [σ(r) ∇Φ(r, t)] = −I_v(r, t)

where Φ(r, t) is scalar electric potential, I_v(r, t) is volumetric current-source density, and σ(r) is the tissue conductivity tensor across concentric cranial layers:

• Brain Parenchyma: Gray matter σ_gray ≈ 0.33 S/m; anisotropic white matter σ_parallel ≈ 0.60 S/m along fiber tracts versus σ_perp ≈ 0.12 S/m transversely.
• Cerebrospinal Fluid (CSF): Conductive subarachnoid fluid (σ_CSF ≈ 1.79 S/m at 37 °C) that shunts tangential currents laterally.
• Cranial Bone: Resistive three-layered skull (σ_skull ≈ 0.004 to 0.015 S/m, ~20 to 80 times less conductive than cortex), strongly attenuating high spatial frequencies.
• Scalp Soft Tissue: Conductive skin and galea (σ_scalp ≈ 0.43 S/m), spreading exiting currents across the outer skull table.

This conductivity sandwich acts as a spatial low-pass filter, smearing focal cortical activity across 2 to 5 cm on the scalp. Unlike Magnetoencephalography—which selectively detects tangential sulcal currents without skull distortion—EEG captures both radial gyral dipoles and tangential sulcal dipoles, though opposing walls of symmetrically active sulci partially cancel.

International 10–20 and High-Density Geodesic Systems

In 1958, Herbert H. Jasper standardized the International 10–20 System, placing electrodes at 10% and 20% fractional arc increments between four cranial landmarks: the nasion, inion, and left and right preauricular points. Labels combine lobar letters (Fp, F, C, T, P, O) with odd left-hemisphere indices, even right-hemisphere indices, and midline z (19–21 clinical channels), extended in research to the 10–10 (64–85 sites) and 10–5 / geodesic systems (128–256 sites at 1.5–2.5 cm pitch).

Clinical montages use sintered non-polarizable Ag/AgCl electrodes with chloride gel (<5–10 kΩ skin impedance) routed into differential instrumentation amplifiers (>1 GΩ input impedance, >100 dB common-mode rejection, 24-bit digitization). Channels recorded against a physical reference (mastoid or Cz) can be transformed into a common average reference or reference-free surface Laplacian (∇²_sΦ, estimating radial scalp current density to sharpen focal topography to ~1.5–2 cm).

Canonical Oscillatory Rhythms and Spectral Decomposition

Voltage traces are decomposed via Welch periodograms, Morlet wavelets, or the Fast Fourier Transform into canonical clinical and cognitive frequency bands:

Above ~30 Hz, frontalis and temporalis Electromyography and microsaccadic ocular spikes exceed cortical gamma power on the scalp, making invasive Electrocorticography preferable for high-gamma (>70 Hz) population tracking.

Sensorimotor Event-Related Desynchronization, P300, and SSVEPs

Requiring zero surgery, scalp EEG underpins three primary non-invasive Brain–Computer Interface control paradigms:

1. Sensorimotor Rhythm (SMR) Motor Imagery: Imagining hand, foot, or tongue movement triggers contralateral event-related desynchronization (ERD)—a 30%–70% power decrease in Rolandic mu (8–12 Hz) and beta (13–30 Hz) bands over C3, C4, or Cz—followed by post-imagery beta rebound (ERS; Pfurtscheller & Lopes da Silva, 1999; Wolpaw et al., 2002), decoded via Common Spatial Patterns (CSP) or Riemannian covariance classifiers.
2. P300 Event-Related Potentials: In oddball character-matrix spellers (Farwell & Donchin, 1988), attending to a target symbol elicits a centro-parietal positive wave 250 to 450 ms post-flash (P300), supporting communication at 5 to 15 characters/min.
3. Steady-State Visually Evoked Potentials (SSVEPs): Focusing on targets flickering at distinct rates (6–40 Hz) entrains phase-locked harmonics over occipital cortex (O1, Oz), decoded by filter-bank Canonical Correlation Analysis at 100 to 300 bits/min.

Minimum-Norm Estimates, LORETA, and Spatial Beamforming

Reconstructing cortical dipole currents J from M scalp potentials V = L J + n across N >> M gray-matter sources (L being the MRI-derived boundary- or finite-element lead-field matrix) poses an ill-posed Helmholtz inverse problem (1853) with infinitely many silent internal solutions (Baillet et al., 2001). Regularization priors yield unique anatomical estimates:

• Minimum-Norm Estimation (MNE) and LORETA: L₂ Tikhonov regularization minimizes ||V − L J||²₂ + λ ||W J||²₂; noise-normalized standardized Low-Resolution Electromagnetic Tomography (sLORETA) and eLORETA remove depth bias to localize smooth distributed cortical sources at ~1–2 cm resolution.
• Linearly Constrained Minimum Variance (LCMV) Beamforming: Data-adaptive spatial filters pass unit gain from a target voxel while minimizing output variance w^T C_V w, rejecting ocular and myogenic interference. Portable caps often co-register EEG electrodes with Functional Near-Infrared Spectroscopy optodes to fuse millisecond electrical dynamics with localized hemodynamic oxygenation.

Why cannot scalp EEG reliably detect isolated single-neuron action potentials?

Single somatic action potentials last ~1 ms and produce quadrupolar extracellular fields decaying as 1/r³, falling below amplifier thermal noise within ~150 μm. Separated by 15–25 mm and resistive skull bone, scalp sensors detect only slower (10–150 ms) postsynaptic potentials summed synchronously across millions of aligned pyramidal dendrites.

Why does closing the eyes markedly increase posterior alpha wave amplitude?

Without visual input during relaxed wakefulness, reciprocal thalamocortical loops between thalamic relay nuclei and occipital cortex enter synchronized 8–13 Hz rhythmic gating (the Berger effect). Visual input or attention depolarizes thalamic cells into tonic transmission, desynchronizing alpha oscillations within 200 ms.

How does high-density 128- or 256-channel EEG improve source localization despite skull blurring?

Standard 19-channel montages spatially undersample scalp voltage gradients. Sampling every 1.5–2.5 cm satisfies the scalp spatial Nyquist limit and, combined with individual T1/T2 MRI finite-element skull models, lowers cortical dipole localization error below 10 mm.

1. Berger, H. (1929). Über das Elektrenkephalogramm des Menschen. Archiv für Psychiatrie und Nervenkrankheiten, 87(1), 527–570. https://doi.org/10.1007/BF01797193
2. Pfurtscheller, G., & Lopes da Silva, F. H. (1999). Event-related EEG/MEG synchronization and desynchronization: basic principles. Clinical Neurophysiology, 110(11), 1842–1857. https://doi.org/10.1016/S1388-2457%2899%2900141-8
3. Baillet, S., Mosher, J. C., & Leahy, R. M. (2001). Electromagnetic brain mapping. IEEE Signal Processing Magazine, 18(6), 14–30. https://doi.org/10.1109/79.962275
4. Wolpaw, J. R., Birbaumer, N., McFarland, D. J., Pfurtscheller, G., & Vaughan, T. M. (2002). Brain–computer interfaces for communication and control. Clinical Neurophysiology, 113(6), 767–791. https://doi.org/10.1016/S1388-2457%2802%2900057-3
5. Buzsáki, G., Anastassiou, C. A., & Koch, C. (2012). The origin of extracellular fields and currents—EEG, ECoG, LFP and spikes. Nature Reviews Neuroscience, 13(6), 407–420. https://doi.org/10.1038/nrn3241