Turning thoughts into text with a new mini brain chip

A tiny 2.5mm2 chip for the human brain could soon improve the lives of people with severe motor disorders. The miniaturised brain-computer interface (MiBMI), developed by a group of researchers at the EPFL in Lausanne, is particularly impressive for its small size.

According to the press release, such systems are usually bulky, consume a lot of energy and are limited in their practical application. The MiBMI represents a major advance in brain-machine interfaces. The chip translates human brain activity into 31 characters with an accuracy of over 93%, a significant improvement over previous versions of such interfaces.

Current brain-machine interfaces record data from electrodes implanted in the brain and then send the signals to a separate computer for decoding. The MiBMI chips record the data but also process the information in real time, speeding up the process significantly.

To enable the interface to decode thoughts, electrodes are implanted in a person's brain. The person imagines writing letters or words, which generates neural signals. The electrodes record the neural activity associated with the motor actions of writing and send it to the MiBMI chip for decoding. The miniature chip processes these signals in real time and translates the brain's intended hand movements into the corresponding digitally readable text. This enables people with severe motor impairments caused by conditions such as ALS, locked-in syndrome or spinal cord injuries to express themselves in writing.

In order to keep the amount of data as small as possible and thus enable the chip to be miniaturised, the researchers had to develop a completely new method of analysing the data from the thoughts. They found that each letter triggers special markers in the brain called distinctive neural codes (DNCs). The chip processes these DNCs – which amount to about 100 bytes, rather than the thousands of bytes of data generated for each letter.

The new chip will eventually be used for more than just text generation. According to the researchers, partnerships are already in place with other research groups for applications in speech decoding or motion control.