EEG, ECoG, or Intracortical: How BCI Signal Acquisition Methods Trade Surgery for Precision

Brain-computer interfaces (BCIs) navigate a core trade-off between invasiveness and signal quality, with methods like EEG, ECoG, and intracortical arrays offering increasing precision at the cost of surgical intervention. The choice of BCI acquisition method hinges on balancing application requirements for signal fidelity against acceptable levels of surgical risk.

Juliana Campos

November 1, 2026 · 4 min read

The Core Trade-off: Invasiveness vs. Signal Quality

The relationship is consistent across the literature: the closer an electrode sits to the neurons it records, the better the signal. EEG electrodes sit on the scalp, where the electrical activity they capture has already been attenuated and distorted by the skull and scalp tissue. The result is a lower signal-to-noise ratio and significant susceptibility to artifacts from muscle and eye movements, which means substantial signal processing is needed to extract usable control information.

Move the electrodes onto the exposed surface of the brain—as ECoG does—and the picture changes. A review of BCI technologies notes that ECoG offers higher spatial resolution, greater signal amplitude, and reduced susceptibility to muscle and eye-movement artifacts compared with EEG. ECoG captures synchronized postsynaptic potential activity from cortical pyramidal neurons directly from the cortex surface, though it requires a craniotomy without penetrating brain tissue itself.

Intracortical arrays go further still, recording from within the brain tissue itself. A BCI review published in PMC observes that motor recovery with non-invasive approaches has been limited by the need for higher-resolution brain signals, and that most researchers agree movement restoration through prostheses with multiple degrees of freedom can only be achieved through invasive approaches. The same review adds that it is unlikely the power of non-invasive modalities will be enhanced in the near future—meaning the trade-off, not a technical workaround, defines the field.

BCI Signal Acquisition Methods: A Comparative Overview

The table below contrasts the three primary electrical methods across the dimensions that matter for technology selection. Spatial resolution figures come from a comparative table of neuroimaging methods in a PMC-published BCI review; the intracortical range spans different signal types, from local field potentials (~0.5 mm) to multi-unit (~0.1 mm) and single-unit activity (~0.05 mm).

Primary electrical BCI signal acquisition methods compared
Method	Invasiveness	Signal Source	Spatial Resolution	Signal-to-Noise Ratio	Typical Applications
Electroencephalography (EEG)	Non-invasive; scalp electrodes requiring good contact and preparation	Summed cortical electrical activity, attenuated and distorted by skull and scalp	Lower; roughly 10 mm, limiting precise localization	Lower, due to signal attenuation and external interference; requires significant processing	Consumer wearables, neuromarketing, basic research; less suitable for high-precision control
Electrocorticography (ECoG)	Semi-invasive; requires craniotomy but does not penetrate brain tissue	Direct recording from the cortex surface—synchronized postsynaptic potentials of cortical pyramidal neurons	Higher than EEG; roughly 1 mm, with micro-ECoG arrays offering greater electrode density	Higher than EEG, with reduced susceptibility to muscle and eye-movement artifacts	Clinical settings such as pre-surgical epilepsy planning; precise cortical mapping
Intracortical arrays	Invasive; electrodes penetrate brain tissue	Direct recording of local field potentials and individual or small groups of neurons	Highest; roughly 0.5 mm (LFP) down to ~0.05 mm (single-unit activity)	Highest signal fidelity of the three methods	High-precision applications such as multi-degree-of-freedom prosthetic control

Two nuances complicate the simple "closer is better" story. ECoG occupies a deliberate middle ground: it offers higher spatial resolution, better signal-to-noise ratio, a wider frequency range, and less training than scalp EEG, while—according to the same comparison—carrying lower technical difficulty and clinical risk than intracortical single-neuron recording, and possibly superior long-term stability. And miniaturized micro-ECoG devices, as described in a review of implantable ECoG technology, add greater-density signal acquisition in a minimally invasive fashion, with increased spatial resolution useful for both diagnosis and BCI research.

Whatever the acquisition method, the downstream pipeline is conceptually shared: raw signals are filtered, features are extracted from them, dimensionality is reduced, and a classifier or decoder translates the resulting feature vectors into device commands. This is why signal quality matters so much—decoding algorithms can only work with what the electrodes deliver. Noisy, low-resolution EEG input demands heavier processing and more user training, while high-fidelity invasive signals support rapid control with minimal training.

Choosing the Right BCI: Applications and Considerations

The market reflects the trade-off. An industry analysis reports that EEG-based BCI applications hold roughly 85% market share, driven by their non-invasive nature, lower cost, and established presence. That dominance makes sense for consumer wearables, neuromarketing, and basic research, where convenience and safety outweigh precision, and where EEG's limitations—lower resolution, artifact susceptibility, and heavy processing demands—are acceptable.

Clinical and restorative applications invert the calculus. ECoG's role in pre-surgical planning for epilepsy patients demonstrates its value where precise localization of brain activity is required, and its balance of fidelity and risk makes it attractive for BCI use in patients already undergoing invasive monitoring. For restoring movement, the PMC review's conclusion is blunt: prostheses with multiple degrees of freedom are generally considered achievable only through invasive approaches.

Notably, the field's boundaries are also practical ones. This comparison covers the three main electrical methods; other modalities such as fNIRS and MEG exist but measure different signals under different constraints. And decoding remains a conceptual, not mathematical, differentiator here—the algorithms matter, but they inherit the quality of the acquired signal.