Stereo-EEG · Grey Matter

Stereo-EEG records the brain's electrical activity from thin electrodes pushed through small holes in the skull into deep structures, and it is used to find where seizures start when the evidence for epilepsy surgery does not agree on a single place.

Stereo-EEG. Stereo-EEG records the brain's electrical activity from thin electrodes pushed through small holes in the skull into deep structures, and it is used to find where seizures start when the evidence for epilepsy surgery does not agree on a single place.

Each electrode is a wire about 0.8 mm thick carrying between 4 and 18 contacts along its length, so one trajectory samples the cortex at its entry, the white matter it crosses and the deep target at its tip. A plan of about ten to fifteen trajectories is drawn on the patient's MRI and placed with a stereotactic frame or a robot; the person then stays on a monitoring ward for days or weeks until enough seizures have been recorded. The method was born at Sainte-Anne Hospital in Paris, where Jean Talairach and Jean Bancaud implanted the first depth electrodes for epilepsy in 1957 and named the technique in 1962.

It reaches what a grid cannot. Electrocorticography covers a patch of surface; depth electrodes reach the hippocampus, the insula and the bottom of sulci, in both hemispheres, without opening the skull.

It samples sparsely. Each contact sees a few millimetres around it, so a seizure that starts between trajectories can be missed, and the plan depends on a good hypothesis.

It is relatively safe. A 2016 meta-analysis found complications in about 1.3 % of patients, haemorrhage and infection being the most common.

It has become a research source. Patients awake for days with electrodes in language areas can do tasks between seizures; in 2025 a speech decoder trained across 25 such patients outperformed decoders trained on each person alone.

Stereo-EEG trades coverage for depth.

A few dozen thin wires can reach almost any structure in the brain, but only along the lines a team chose in advance.

Questions: When do epilepsy teams choose depth electrodes over a grid on the surface of the cortex? A grid needs a large opening in the skull and sees densely the surface it lies on; depth electrodes go in through small holes and sample sparse points along lines that can reach deep structures in both hemispheres. Teams choose depth electrodes when the suspected onset is deep (the hippocampus, the insula, the bottom of a sulcus), when it could lie on either side, or when several distant regions are candidates. A grid keeps its place when the focus is near the surface and the cortex around it must be mapped finely, next to language or motor areas for example. Many centres that once relied on grids now start with depth electrodes, partly because their complication rate is lower. How do depth electrodes tell which hippocampus starts the seizures? Scalp EEG sees a temporal lobe seizure only once it has spread toward the surface, and by then both sides may be involved, so it can point to the wrong hemisphere or to both. With an electrode in each hippocampus and others in the neighbouring cortex, the recording shows where the first change appears, often a burst of fast low-voltage activity at one contact, and how many seconds later it reaches the other side. If every recorded seizure starts in the same hippocampus, removing or ablating it has a good chance of stopping them; if they start on both sides, surgery on one side is unlikely to help and a stimulator is considered instead. Why do so many speech-decoding studies use recordings from people with epilepsy? They already have electrodes inside the skull for clinical reasons and spend days on a ward waiting for seizures, and many agree to do research tasks in between. Their electrodes are placed to find seizures, so each person covers a different, uneven set of regions, which used to confine every decoder to one patient. A 2025 study in Nature Communications turned that into a strength: it pooled stereo-EEG recordings from 25 people saying tongue twisters aloud, and the decoder trained on the group predicted phonemes better than decoders trained on each person alone. These patients can still speak, so the work maps the speech network and tests methods, while a speech neuroprosthesis for paralysis still needs its own implant.