MIT Researchers Think They Can Spot Early Signs of Parkinson’s in the Way People Type

By monitoring how long we hold down keystrokes, it may be possible to detect neurological diseases years before other symptoms appear

© Andrew Brookes/Corbis

From the physical keys on our laptops to the software buttons on our smartphones, most of us rely on keyboards as the primary way of entering data into the digital world. But it turns out that our keyboards can also tell us quite a bit about ourselves, detecting when we’re tired, drunk, and even when we’re showing early signs of neurological disorders like Parkinson’s disease—perhaps years before more recognizable symptoms surface.

Researchers at the Madrid-MIT M+Visión Consortium, a network dedicating to healthcare innovation in Madrid, are gathering and analyzing the keystrokes of volunteers with software and studying the patterns that emerge through machine learning. Individual typing patterns have already been used to identify individuals; some banks have used them to increase security when logging into accounts. But according to a soon-to-be-published paper in Scientific Reports, the M+Visión team was able to take the same typing data, combined with pattern recognition techniques, to distinguish between typing done when fully rested and when volunteers were tasked to type when woken up in the night. That data could also be used to detect neurological conditions much earlier than existing methods.

Parkinson's diagnosis by typing on a keyboard

To be clear, the team is only gathering information about the timing of key presses, not which keys are being pressed. The researchers developed software that could be applied to a web browser to track how long a typist holds down each key. There’s no need to use specialized keyboards, and little cause for privacy concerns. In fact, many third-party smartphone keyboards gather much more data about what we type.

But it’s clear from the group’s work that we leave behind a trove of information when we interact with electronic devices in our daily lives.

“Every time we touch something that has a microprocessor in it, the microprocessor is able to measure the timing with sub-millisecond accuracy,” Luca Giancardo, a M+Vision fellow and the paper’s first author says. “You can get potential information from a microwave, but changing the software in a microwave is much harder.”

The paper primarily focuses on recognizing fatigue, as that’s one of the most common forms of motor impairment. A group of volunteers first typed a Wikipedia article during the day and then were asked to type another article after being woken up 70 to 80 minutes after going to sleep; in the latter scenario, the timing of their keystrokes was more inconsistent. But according to MIT, a preliminary study involving 21 volunteers with Parkinson’s and 15 people without the disease indicated that those with Parkinson’s show more keystroke variation.

“There is a motor decline seven years before clinical diagnosis [is possible], and the motor decline carries on,” says Giancardo. He says catching signs of the disease earlier would allow neurologists to tweak treatment based the patient’s motor decline, and perhaps eventually halt the decline early on with treatments that are currently in development.

The technique might eventually be used to test for other neurological diseases, as well as rheumatoid arthritis, and whether or not the person typing is drunk. For now though, the team is focused on proving, improving and refining their method for detecting Parkinson’s with a larger study.

Beyond that, the researchers are also interested in gathering a larger swathe of keyboard input from a broad group of users, which should give them a better typing pattern baseline and help them diagnose different conditions.

“Hopefully we’ll be able to partner with some big players, so our technology can be included on larger platforms, and the signal can be captured without user intervention," says Giancardo. “They would just have to either opt out or opt in.”

Until that happens, the team is doing some crowdsourcing of data on their own. They’ve developed an app, available at, which monitors typing in Windows or Mac OSX in much the same way as their controlled studies. Healthy typists can share their keyboard data, and users that have been diagnosed with Parkinson’s can indicate that when signing up, as well as the stage of their illness and what medications they’re taking.

Get the latest stories in your inbox every weekday.