Tech, ExplainedHealthtech and life sciences
Product

Non-invasive brain-computer interface (BCI)

Reading the brain without surgery.

BEFORE ITAFTER ITNon-invasivebrain-computerinterface (BCI)NexstemCOVERSEEG headsetSignal cleaningOn-device AINo surgerysendsstress scoresDigital therapeutics formental healthruns themodelon deviceEdge AI↗ GENERATIVE AI APPLICATIONSA user puts on the headset,and its electrodes touchthe scalp.BEFORE ITAFTER ITNon-invasivebrain-computerinterface (BCI)NexstemCOVERSEEG headsetSignalcleaningOn-device AINo surgerysendsstress scoresDigitaltherapeutics formental healthruns themodelon deviceEdge AI↗ GENERATIVEAI APPLICATIONSA user puts onthe headset,and itselectrodestouchthe scalp.BEFORE ITA user puts on the headset, and itselectrodes touch the scalp.↓Non-invasive brain-computerinterface (BCI)NexstemEEG headsetSignal cleaningOn-device AINo surgeryAFTER IT↓ sends stress scoresDigital therapeutics for mental health↓ runs the model on deviceEdge AI↗ GENERATIVE AI APPLICATIONS
The non-invasive BCI headset cleans the brain's EEG signals, runs a model on the device with edge AI, and sends stress scores to a mental health therapeutics app.
opens a page inside this one a page in the same scene a page in another scene a concept page a label only, with no page

A non-invasive brain-computer interface (BCI) is a headset that reads the brain's electrical signals from outside the head and turns them into measurements or commands for a computer, with no surgery and no implant. Most use EEG, electroencephalography. Small sensors called electrodes touch the scalp. They pick up the tiny voltages, a few millionths of a volt, that brain cells produce when many of them are active together.

There are two kinds of BCI. An invasive BCI places electrodes inside the skull, through surgery. It picks up clearer signals, but every user needs an operation. A non-invasive BCI sits on the head like a cap or a headband. Its signals are weaker and noisier, because they pass through the skull and the scalp. Blinks, jaw movements and the electrical wiring in the room add noise too. So much of a non-invasive BCI's work is done by the software that cleans the signal.

Here is how the signal becomes useful. A headset with 8 electrodes measures each one 250 times a second, which is 2,000 numbers every second. Software removes the noise from blinks and wiring. It then splits the signal into bands by frequency. For instance, alpha waves, at 8 to 12 cycles a second, are stronger when a person is relaxed with closed eyes. A model trained on labelled recordings turns these patterns into an output: a stress score, a sign of drowsiness, or a choice between two commands. More and more often, this model runs on the headset or the phone itself, which is called edge AI. Then the raw brain signals do not have to leave the device.

Take, for example, a research lab at an engineering college in Bengaluru that studies how attention changes during long tasks. Twenty volunteers wear headsets while they do a 40-minute task on a computer. The software records their EEG, marks the moments when their attention drops, and saves the data with the time of each reading, as time series data, for analysis.

The same headsets are used in health products. A stress headband in digital therapeutics for mental health uses EEG to train a person to relax. Research groups also use EEG to help people control a wheelchair or a cursor after paralysis, though implants are usually more reliable for that. Brain signals are sensitive personal data, which India's data protection law (DPDP Act) protects.

Tech, Explained is a part of ingrid.fyi. Every page explains one piece of technology in plain words.