Non-invasive Brain-Computer Interface

When nerve cells in the brain are active, tiny electrical currents flow. These currents generate voltages that can be measured on the scalp. A non-invasive brain-computer interface does exactly that: it picks up such signals using sensors from outside and sends them to a computer. The computer searches the measurements for recognizable patterns and turns them into a command, such as moving a cursor. “Non-invasive” means that nothing is operated on or placed under the skin. You put on a cap or headband and take it off again afterward.

The appeal of an interface without a scalpel

The alternative is invasive systems, in which electrodes are implanted into the brain or beneath the skull. Such procedures are surgeries with real risks: bleeding, infections, scar tissue. That’s why, so far, only a small number of severely ill people receive them. A non-invasive system, by contrast, can be tried by anyone, including healthy test subjects in a research lab.

This is what makes the technology broadly researchable and marketable in the first place. Depending on quality, an electrode cap costs anywhere from a few hundred to a few thousand euros. It requires no clinic, no surgeon, and no months-long approval process. Companies can use it to test products on many people instead of a handful of patients.

Still, the most important field of application remains medicine. People recovering from a stroke train movements using such systems: they imagine a hand movement, and a screen or a splint responds accordingly. This direct feedback helps the brain relearn. Even for people who can barely move at all, simple yes-no signals are already a major gain.

From voltage pattern to mouse click

The most common measurement method is EEG, short for electroencephalography. Here, metal electrodes rest on the scalp and measure voltage differences in the range of millionths of a volt. There are also methods that use infrared light to measure blood flow in the brain, as well as large magnetic measuring devices found in clinics. All of them share the same fundamental problem: between the nerve cell and the sensor lie the meninges, the skull, and the skin.

A good analogy is a football stadium. From outside, you can hear whether a goal has just been scored, but not what a single spectator is shouting. In the same way, EEG averages the activity of millions of cells into one coarse overall signal. Invasive electrodes, on the other hand, sit right in the middle of the stands and can pick out individual voices. That’s why invasive systems today can control robotic arms, while non-invasive ones usually manage only simpler tasks.

To turn the noise into a command, machine learning comes into play. The system is first calibrated: the person repeatedly imagines a particular movement, and a program learns what the corresponding pattern looks like. After that, it can recognize this pattern. Calibration often takes 10 to 30 minutes and frequently has to be repeated, since the signals change from day to day.

Between rehab clinic and gaming convention

In the news, non-invasive BCIs often show up in comparison with companies like Neuralink. Their implants are invasive and make headlines. The non-invasive variant is then described as the safer but slower alternative. Anyone reading such reports should always check whether a single person was measured in a lab under ideal conditions, or whether the device actually works in everyday life.

Retailers already sell headbands for meditation, concentration training, or sleep analysis. These devices usually measure only a few crude metrics and are far from being mind readers. A common misconception is that such sensors could read out thoughts or sentences. They recognize states and trained patterns, not content.

Still, the data question should be taken seriously. Brain data is particularly sensitive, and some countries are already working on their own regulations for it. Anyone using such a device should know where the measurement data is being sent. For the coming years, the following holds true: the technology has arrived in rehabilitation, but not yet as an everyday form of control for everyone.