Electroencephalography

Nerve cells in the brain communicate using tiny electrical voltages. When many cells become active at the same time, this adds up to a signal that reaches all the way to the scalp. Electroencephalography measures exactly this signal. For this purpose, small metal plates, called electrodes, are attached to the outside of the head. The measured voltages are extremely small, usually a few millionths of a volt, and are therefore heavily amplified. The result is a curve showing how brain activity changes from millisecond to millisecond.

What an EEG delivers for medicine and research

The EEG is the oldest method that allows us to watch a living brain at work. The German psychiatrist Hans Berger performed the first measurement on a human in 1924. To this day, the method is inexpensive, harmless, and can be used practically anywhere. A device costs a fraction of an MRI scanner and requires no dedicated room.

In medicine, the EEG is virtually irreplaceable for epilepsy. An epileptic seizure occurs when large groups of cells fire uncontrollably in synchrony. This exact pattern becomes visible in the curve, often even between seizures. Sleep labs also work with it, because sleep stages can be recognized by typical wave forms. Even the question of whether activity is still present after severe brain damage is answered using EEG.

For the tech industry, the EEG is interesting because it is the simplest route to a brain-computer interface. This refers to a connection in which a computer reacts directly to brain signals, without keyboard or speech. These signals are the raw material that learning programs later analyze.

From electrode to wave curve

In a clinical measurement, typically 19 to 64 electrodes are distributed over the head according to a fixed scheme. A conductive gel ensures good contact between the metal and the skin. Each electrode does not measure an absolute value, but always the voltage difference to a reference point, often at the ear. An amplifier boosts the tiny signal, after which a converter turns it into numbers. Common sampling rates are 250 to 1000 measurements per second per electrode.

The raw data is almost always contaminated. Eye blinks, chewing movements, and the 50-hertz hum from the power outlet overlay the brain signals. Such disturbances are called artifacts and must be filtered out. The curve is then decomposed by frequency: alpha waves around 10 oscillations per second occur during relaxed wakefulness, slow delta waves during deep sleep.

The major weakness of the EEG is its spatial resolution. Think of it like listening from outside a stadium wall: you can hear exactly when there is cheering, but not which row of seats it came from. Skull bone and skin blur the signals, so the source within the brain can only be roughly estimated. In return, the temporal resolution is excellent, far better than with an MRI. This is precisely where machine learning programs come in, filtering out patterns from noisy curves that a human would overlook.

EEG in clinics, headsets, and headlines

Most commonly, one encounters the EEG in a hospital or at the neurologist’s office. The examination usually takes 20 to 30 minutes, does not hurt, and leaves nothing behind but sticky hair. Anyone who has ever spent a night in a sleep lab also wore electrodes on their head.

In addition, consumer EEG headsets have been available for a few years, with a small number of dry electrodes and prices starting at around 200 euros. They advertise meditation feedback, concentration training, or control by thought. Signal quality is well below that of clinical devices, and many promises do not hold up to scrutiny. A common misconception is that an EEG can read thoughts. It detects states and rough patterns, not sentences.

In the news, the EEG appears mainly in connection with brain-computer interfaces. Unlike implants, which are surgically placed inside the brain, the EEG works from the outside and without any intervention. Paralyzed individuals already use it in studies to control cursors or letter boards. Because such data is highly personal, legal scholars are now discussing separate protection rules for brain data.