Electrocorticography, commonly abbreviated as ECoG, is a technique that records electrical brain activity by placing electrodes directly on the surface of the cerebral cortex. Because the electrodes sit beneath the skull rather than on the scalp, ECoG captures neural signals with far greater spatial detail and signal clarity than a standard EEG. For decades its primary home has been epilepsy surgery, where neurosurgeons use it to locate seizure-generating tissue before removing it. But ECoG has increasingly become a platform for brain-computer interfaces, speech decoding, and closed-loop stimulation devices, making it one of the more versatile tools in modern neuroscience and neurosurgery.
What Makes ECoG Different from a Scalp EEG
A conventional EEG sticks electrodes to the outside of your head. The skull, scalp, and the fluid layers between them act as biological filters, blurring and weakening neural signals before they ever reach the sensor. ECoG bypasses all of that. Electrodes rest directly on the cortical surface, sometimes on top of the outermost protective membrane (epidural) and sometimes beneath it (subdural). This proximity means ECoG can resolve activity from patches of cortex only a few millimeters across and can detect high-frequency oscillations, especially in the gamma band above roughly 30 Hz, that scalp EEG almost never picks up cleanly.
A study that compared ECoG, EEG, and fMRI while subjects watched the same movie clip found that the fMRI blood-flow signal correlated positively with high-frequency ECoG power across task-related brain areas, while it correlated with low-frequency power in scalp EEG and ECoG alike. The finding underscores that ECoG can simultaneously track both the slow, large-scale rhythms visible on scalp recordings and the fast, focal activity that only direct cortical contact reveals.1PubMed Central / Neuroimage. Elucidating relations between fMRI, ECoG, and EEG through a common natural stimulus
Subdural Grids Versus Depth Electrodes
When a patient with drug-resistant epilepsy needs invasive monitoring, clinicians choose between two main electrode approaches. Subdural electrodes (SDE) are thin, flexible sheets or strips laid directly on the brain surface through a craniotomy. Stereoelectroencephalography (SEEG) instead threads thin depth electrodes through small burr holes into the brain, sampling activity at multiple points along each shaft. Both yield ECoG-quality signals, but from different vantage points: subdural arrays cover broad cortical surfaces while depth electrodes probe deeper structures and both hemispheres without a large opening.
The shift toward SEEG has been one of the bigger trends in epilepsy surgery over the past decade. A meta-analysis pooling 16 studies and roughly 3,750 patients found no statistically significant difference between the two approaches in seizure freedom after surgery, but SEEG was associated with about half the complication rate of subdural grids, fewer major bleeding events, fewer neurological deficits after the procedure, and operative times that were more than an hour shorter on average.2PubMed. Stereoelectroencephalography versus subdural electrodes for invasive monitoring of drug-resistant epilepsy patients: a systematic review and meta-analysis A second meta-analysis came to a slightly more favorable conclusion for SEEG on seizure outcomes, finding a modest but statistically significant improvement in favorable seizure outcome rates along with roughly the same halving of complications.3Journal of Neurosurgery. Comparative assessment of stereoelectroencephalography and subdural electrodes in invasive epilepsy monitoring: a systematic review and meta-analysis A large propensity-matched U.S. cohort analysis showed patients who had subdural grids were more likely to proceed to resective surgery but also had higher complication rates, while SEEG-guided resections were associated with higher odds of seizure freedom.4PubMed Central. Comparative Effectiveness of Stereo-EEG versus Subdural Grids in Epilepsy Surgery
None of this means subdural grids are obsolete. They remain particularly useful when the suspected seizure focus is on a broad cortical surface and when the surgical team wants to combine seizure mapping with functional cortical mapping in the same monitoring session. The choice between the two is case-specific, driven by where the seizure focus is suspected and what information the team needs before operating.
ECoG in Epilepsy Surgery
Epilepsy surgery has been ECoG’s core clinical application since the mid-twentieth century. The pioneering work of Wilder Penfield and Herbert Jasper at the Montreal Neurological Institute established the basic template: record directly from the brain to find pathological activity, stimulate the cortex to identify functional areas, then remove as much of the seizure-generating tissue as possible while sparing eloquent cortex.5Seizure. The history of invasive EEG evaluation in epilepsy patients
Intraoperative ECoG, performed while the patient is in the operating room, continues to serve this role in specific scenarios. It is most useful for tailored temporal lobe surgeries and for guiding removal of developmental brain malformations, especially cortical dysplasias, where the borders of abnormal tissue can be electrically distinct from surrounding cortex. For the most straightforward form of temporal lobe epilepsy, where MRI already shows clear structural changes, intraoperative ECoG adds little because the surgical plan is already well defined by imaging. But when a surgery targets cortical dysplasia, ECoG is often the main tool guiding how much tissue to remove, and residual abnormal spiking after resection has been linked to a higher chance of seizures returning afterward.6PubMed. Intraoperative electrocorticography in epilepsy surgery: useful or not?
Mapping Language and Motor Areas
Beyond finding seizure sources, ECoG is used to map functional brain areas so surgeons know what not to cut. The traditional gold standard is electrical cortical stimulation (ECS), where a mild current is applied to pairs of electrodes one at a time while the patient performs tasks like naming objects. If stimulation at a particular spot disrupts speech or triggers involuntary movement, that area is flagged as essential.
ECS works, but it is slow and carries a risk of triggering seizures during testing. An alternative approach records high-gamma activity, the fast electrical oscillations above about 70 Hz, from the ECoG grid while the patient performs the same tasks. Sites that light up with high-gamma power during naming tend to overlap with sites that ECS identifies as language-essential. A meta-analysis comparing the two methods found that high-gamma mapping is specific but not very sensitive relative to ECS, meaning that when it flags a site, the site is very likely to matter, but it misses some sites that ECS would have caught.7PubMed Central. ECoG high-gamma modulation versus electrical stimulation for presurgical language mapping An earlier study put the specificity at about 84% when combining naming and motor tasks, with sensitivity around 43%.8PubMed. Electrocorticographic high gamma activity versus electrical cortical stimulation mapping of naming
The practical upshot is that high-gamma mapping can be recorded passively from all electrodes at once, without any stimulation, while ECS must be done sequentially at one electrode pair at a time. That speed advantage makes ECoG gamma mapping a useful first pass: it generates a preliminary functional map that tells the surgical team which electrode sites are low priority for the more time-consuming ECS testing. In children, this advantage is even more pronounced because pediatric patients struggle with awake craniotomies, and their immature cortex sometimes requires higher stimulation intensities that increase the risk of intraoperative seizures.9Journal of Clinical Neurophysiology. Pediatric Intraoperative Neurophysiologic Mapping and Monitoring in Brain Surgery A study of high-gamma mapping during spontaneous conversation in children found sensitivity of about 89% and specificity of about 64% compared with ECS, suggesting that in at least some pediatric cases, passive ECoG mapping may perform better than the average adult results would predict.10PubMed. Electrocorticographic language mapping in children by high-gamma synchronization during spontaneous conversation: comparison with conventional electrical cortical stimulation
Closed-Loop Stimulation for Seizure Control
ECoG is not only a diagnostic tool. It also forms the sensing backbone of the RNS System, a commercially available closed-loop neurostimulator for epilepsy. The device sits inside the skull and continuously monitors ECoG signals from leads placed at the seizure focus. When it detects the electrical signature of an oncoming seizure, it delivers a brief pulse of stimulation aimed at stopping it. Controlled trials in the U.S. showed that seizure reduction continued to improve over time, reaching about 75% reduction after nine years of treatment. A unique secondary benefit is that the device logs ambulatory ECoG data around the clock, giving clinicians a window into each patient’s seizure patterns that was previously impossible to obtain outside a hospital monitoring unit.11PubMed Central. Responsive Neurostimulation as a Novel Palliative Option in Epilepsy Surgery
Research into how the device actually works at a neurophysiological level has revealed that the stimulation’s most clinically meaningful effects are indirect. Direct effects like immediately halting an electrical seizure discharge were not reliably linked to better outcomes. Instead, the effects that predicted clinical improvement were more subtle network-level changes happening away from the stimulation site, including spontaneous seizure inhibition and alterations in the frequency patterns of seizure activity over time.12JAMA Neurology. Association of Closed-Loop Brain Stimulation Neurophysiological Features With Seizure Control Among Patients With Focal Epilepsy This suggests the device may be gradually retraining the seizure network rather than simply interrupting individual events.
Brain-Computer Interfaces and Speech Decoding
ECoG occupies a middle ground in the world of brain-computer interfaces (BCIs). Intracortical microelectrode arrays, which penetrate into brain tissue, offer finer-grained access to individual neurons but tend to degrade over months to years as scar tissue builds up around the electrode tips. Scalp EEG is completely noninvasive but too blurry for many BCI tasks. ECoG threads the needle: it records from the cortical surface with enough resolution to decode arm movement direction, individual finger flexion, and even attempted speech, while causing comparatively less tissue disruption than penetrating electrodes.
An early online BCI study demonstrated that four out of five epilepsy patients could use ECoG signals from arm movements to steer a computer cursor in one of two directions, with correct decoding rates averaging 75% of trials.13Journal of Neural Engineering. An online brain–machine interface using decoding of movement direction from the human electrocorticogram Since then, machine learning techniques have dramatically improved what can be extracted from ECoG. A study using gradient-boosted decision trees combined with time-concatenated features achieved about 77% accuracy in classifying which individual finger was moving across a six-class task, outperforming previous deep-learning approaches while training roughly 250 times faster.14Journal of Neural Engineering. Fast and accurate decoding of finger movements from ECoG through Riemannian features and modern machine learning techniques Deep learning models that combine convolutional feature extraction with recurrent networks have also been applied to continuous finger trajectory decoding from ECoG.15PubMed. Decoding of finger trajectory from ECoG using deep learning
The speech domain is where ECoG-based BCIs have generated the most public excitement. Research has shown that the cortical areas controlling vocalization and articulation span a large region of the ventral sensorimotor cortex, and that speech acoustics can be reconstructed from ECoG if electrode arrays are dense enough to sample that region comprehensively.16PubMed. Brain-Computer Interface: Applications to Speech Decoding and Synthesis to Augment Communication These findings underpin ongoing efforts to build neuroprosthetic devices that could restore communication for people who have lost the ability to speak due to conditions like ALS or brainstem stroke.
How Safe Is the Procedure
Any surgery that opens the skull and places hardware on the brain carries risk, and ECoG monitoring is no exception. The specific risks depend heavily on whether subdural grids or depth electrodes are used, how long the electrodes stay implanted, and how many electrodes are placed.
For subdural grids, one series of 50 implantations found major complications in about 16% of cases, including delayed blood collections requiring emergency re-operation in roughly 8% and infection in about 4%.17PubMed. Complications and results of subdural grid electrode implantation in epilepsy surgery A separate series reported an 11% surgical complication rate, though with no permanent morbidity or mortality.18PubMed. Outcome and complications of chronically implanted subdural electrodes for the treatment of medically resistant epilepsy A national-level analysis of invasive ECoG in the United States found that the most common complication was cerebrospinal fluid leak at about 12%, while only one death was identified across the entire study cohort.19PubMed Central. National trends and complication rates for invasive extraoperative electrocorticography in the USA
As noted in the electrode-configuration discussion, the meta-analyses comparing SEEG to subdural grids consistently find that SEEG carries roughly half the complication rate, with substantially fewer bleeding events and neurological deficits. This safety advantage has been a major driver of the field’s ongoing shift toward depth electrodes for diagnostic monitoring.
Can ECoG Electrodes Stay In Long Term
Most clinical ECoG monitoring lasts days to a few weeks. But the ambitions of brain-computer interfaces demand electrodes that function for years. A central question is whether the body’s response to a foreign object sitting on the brain will degrade signal quality over time.
A study that kept a high-density ECoG grid implanted over motor cortex in a macaque for 666 days found minimal direct damage to the underlying brain tissue, though the grid itself became encapsulated in a collagen-rich scar. Despite that encapsulation, the electrodes still picked up movement-related cortical signals more than 18 months after implantation.20PubMed Central. Histological Evaluation of a Chronically-implanted Electrocorticographic Electrode Grid in a Non-human Primate A separate primate study using a wireless implantable ECoG system also found tissue thickening around the electrodes without overt cortical inflammation, though the fibrotic growth did reduce signal amplitude over time.21PubMed Central. Chronic subdural electrocorticography in nonhuman primates by an implantable wireless device for brain-machine interfaces
This signal attenuation from scar tissue is the main engineering challenge for chronic ECoG. One recent approach involves coating electrode arrays with a fibrosis-resistant hydrogel. In rodent testing, hydrogel-coated arrays showed a 20-fold reduction in early impedance increases compared with uncoated arrays, and they retained about 95% of their signal-to-noise ratio over 16 weeks, versus roughly 70% for uncoated arrays.22PubMed Central. Long-Term Stable Subdural Recordings Enabled by Fibrosis-Resistant Hydrogel-Integrated µECoG Arrays Data from two fully implanted wireless ECoG devices in nonhuman primates performing motor tasks over many sessions further demonstrated that cortical signals remained meaningfully correlated with behavior months into the recording period.23Nature. Longitudinal multitask wireless electrocorticography data from two fully implanted nonhuman primates
Endovascular ECoG and the Stentrode
One of the more inventive recent developments sidesteps the skull entirely. The Stentrode is a stent-mounted electrode array threaded through a blood vessel in the neck, up into the superior sagittal sinus, a large vein that runs along the top of the brain. Once in position, its electrodes sit close enough to motor cortex to record ECoG-quality signals without any craniotomy at all.
A feasibility study in four patients with severe paralysis found that the endovascular device appeared safe and was able to record stable signals suitable for long-term BCI use. Endovascular signal quality has been shown to be comparable to that of subdural arrays, though traditional subdural electrodes retain a slight edge in amplitude and spatial resolution.24JAMA Neurology. Assessment of Safety of a Fully Implanted Endovascular Brain-Computer Interface for Severe Paralysis in 4 Patients: The Stentrode With Thought-Controlled Digital Switch (SWITCH) Study A broader review of endovascular neural recording confirmed these findings, noting similar bandwidth, power, and signal-to-noise characteristics between endovascular and epidural or subdural electrodes.25Journal of Neural Engineering. Making a case for endovascular approaches for neural recording and stimulation
The appeal is obvious: a procedure that looks more like placing a cardiac stent than performing brain surgery could dramatically expand who is eligible for an ECoG-based BCI. Patients who might not tolerate or want a craniotomy could receive an endovascular device through a routine catheterization. The trade-off is less flexibility in where the electrodes end up, since they can only reach brain areas that happen to lie adjacent to accessible veins.
Next-Generation Electrode Arrays
Traditional clinical ECoG grids use electrodes a few millimeters in diameter, spaced about a centimeter apart. That spacing is adequate for localizing a seizure focus spanning a lobe of the brain but too coarse for decoding fine motor actions or mapping cortical columns. A newer class of micro-ECoG (μECoG) arrays uses electrodes as small as tens of micrometers and can pack hundreds or thousands of contacts onto a flexible substrate.
An influential design integrated ultrathin silicon transistors directly into the electrode array, allowing thousands of sensors to be amplified and multiplexed on the array itself rather than requiring a separate wire for each electrode. This active-electronics approach keeps the cable bundle manageable even at very high channel counts.26PubMed Central. Flexible, foldable, actively multiplexed, high-density electrode array for mapping brain activity in vivo On the animal research side, a 60-electrode multiplexed μECoG system was developed specifically for recordings in freely moving rats, addressing the practical problem that many-channel arrays with individual wires are too bulky for small animals to carry around.27Journal of Neural Engineering. A low-cost, multiplexed μECoG system for high-density recordings in freely moving rodents
These engineering advances are converging with the chronic-stability improvements discussed earlier. The long-term goal is a fully implantable, wireless, high-density μECoG array that can record from thousands of cortical sites for years without signal degradation. That device does not exist yet, but the individual pieces, flexible substrates, on-array multiplexing, anti-fibrotic coatings, and wireless data transmission, are all in active development.
ECoG as a Window Into Cognition
Epilepsy patients who undergo invasive monitoring spend days in the hospital with electrodes already in place, creating a unique opportunity for cognitive neuroscience research. With informed consent, researchers present these patients with auditory stimuli, memory tasks, or language exercises and record the brain’s response at a resolution that scalp EEG cannot match and fMRI cannot capture in real time.
A review of this research highlights how intracranial EEG, including ECoG, has advanced understanding of auditory prediction, working memory, episodic memory, and internally directed thought, giving human researchers the kind of single-trial precision that was once available only in animal experiments.28Journal of Neural Engineering. Insights into human cognition from intracranial EEG: A review of audition, memory, internal cognition, and causality One concrete example: an ECoG study of auditory processing found that the brain’s response to unexpected sounds, called mismatch responses, was driven primarily by a deviance-detection mechanism in the posterior superior temporal gyrus, with adaptation playing a smaller, location-dependent role.29PubMed Central. Deviance detection is the dominant component of auditory contextual processing in the lateral superior temporal gyrus: A human ECoG study
This line of work is opportunistic by nature. Electrode placement is dictated by clinical need, not research interest, so coverage is always incomplete and biased toward brain regions commonly involved in epilepsy. Still, the accumulated data from thousands of patients over decades has provided insights into human brain function that would be ethically impossible to obtain through elective surgery in healthy volunteers.