Brain-computer Interfaces Decode Handwriting

 

By: Allison Kubo

Image Credit: Nature 593, 249-254(2021) https://www.nature.com/articles/s41586-021-03506-2

The study participant, T5, was paralyzed from the neck down, but it was translated onto the screen when he imagined writing. Brain-computer interfaces (BCI) aim to restore function to those who have difficulty or even lost the ability to move or speak. And, yes, it would allow you to control a computer with your brain. New research published in Nature on May 12 shows progress in restoring writing to those who have suffered paralysis. Study participant T5, given this denomer to maintain their privacy, used BCI type at near the speed of the control group with high accuracy. They achieved 94.1% raw accuracy and greater than 99% accuracy with an autocorrect function.

How does this work? Studies have shown that even after years of paralysis, thinking about the desired movement still activates the necessary parts of the brain. The researchers instructed T5 to “imagine” writing with a pen on lined paper. The brain fired their signals but something in the communication, a spinal cord injury, for example, stopped it from moving to the hand. However, by detecting the brain signals using two microelectrodes positioned on the skull near the brain that controls the right-hand movement (precentral gyrus), the signal can be translated to words by the computer rather than the hand. Of course, just a raw electromagnetic signal from the brain would be impossible to decode. But by training the model using several recordings of known signals, researchers can match the signals to known letters.

Why does this work? This recent study was able to increase typing speeds to 90 words per minute. Compared to the point-and-click typing or methods using eye movements, measured at only 40 characters per minute, this offers a significant improvement. Point-and-click methods have people move a cursor on a screen and detect the signals similarly. Although this method is perhaps more varied since it could be applied to games, websites, etc., it is limited by how fast the computer can decode the detected signals. This is far slower than decoding handwriting signals. One possible explanation is that writing is a far more varied signal than moving a cursor in a straight line. The curves of the letters vary over time but moving a cursor is more constant motion in a single direction.

The method is not yet available widely to those who would need it. But it offers a new option to restore precious communication to those who suffer from spinal cord injuries or diseases such as ALS.

You may also read these articles