Swiss chip accurately translates thoughts into text

A complete brain-machine interface on a chip. CC-BY-SA 4.0© 2024 EPFL

Swiss researchers at the Swiss Federal Institute of Technology Lausanne (EPFL) have developed a microchip that converts thoughts into text with high precision while consuming very little power.

This tiny chip, just a few square millimetres in size, is a miniaturised brain-computer interface (MiBCI) created by a research group at EPFL. The development was reported in the IEEE Journal of Solid-State Circuits.

According to an EPFL press release, the main innovation is its miniaturisation. Typically, such systems are “bulky, energy-intensive, and limited in practical application”. The MiBCI, however, represents a major advancement in brain-machine interfaces. The chip translates human brain activity into 31 characters with an accuracy of 91.3%, a significant improvement over previous versions of similar interfaces.

“MiBCI allows us to convert complex neural activity into readable text with high precision and low power consumption,” says Mahsa Shoaran, head of EPFL’s Integrated Neurotechnologies Laboratory (INL). “This breakthrough brings us closer to practical, implantable solutions that could significantly enhance the communication abilities of individuals with severe motor deficits.”

This could benefit people suffering from conditions such as amyotrophic lateral sclerosis (ALS), locked-in syndrome, or spinal cord injuries, all of whom experience severe motor impairments.

Possible help for people with severe motor deficits. CC-BY-SA 4.0© 2024 EPFL

Recognising imagined handwriting

To enable the interface to decode thoughts, electrodes are implanted in a person’s brain. The person then imagines writing letters or words. This generates neuronal signals, which are recorded by the electrodes. These signals, associated with the motor actions of writing, are sent to a separate computer for decoding. The MiBCI chip processes the signals in real time, translating the intended hand movements into corresponding, digitally readable text.

EPFL describes the development as “a masterpiece of extreme miniaturisation”, combining expertise in integrated circuits, neurotechnology, and artificial intelligence.

Analyse data in other ways

The research team had to come up with a novel approach to data analysis in order to process the large amount of information captured by the electrodes of the miniaturised brain-machine interface. The researchers found that the brain activity for each letter (when the person imagines writing it by hand) contains very specific markers called distinctive neural codes (DNC).

The microchip will only process these DNCs – which amount to about 100 bytes – and not the thousands of bytes of data generated for each letter. In this way, the system is fast, precise and consumes little energy. According to the research team, this advance also enables “faster learning times, making the ICM easier and more accessible to use”.