Wearable Sensor Tracks How Much You Scratch

Data could be a boon for people with eczema and other diseases involving intense itch

3 min read

Infrared camera footage shows a child scratching while wearing an ADAM sensor. The outputs of the sensors are shown as the child scratches.
Infrared camera footage shows a child scratching while wearing an ADAM sensor. The outputs of the sensors are shown as the child scratches.
Image: Shuai Xu

How much does it itch? That’s a question doctors often ask people with eczema and other itch-inducing ailments, but the answer is subjective and hard to quantify; like being asked to rate one’s pain using one of those emoji-face pain scales

Today researchers report in Science Advances a wearable sensor that can accurately quantify how much a person is scratching. Worn on the back of the hand, the gadget uses both motion and vibration sensing to sense the hand’s activity. A machine learning algorithm then identifies the motion as either scratching or some other non-scratching hand motion. 

“The ability to quantify scratching as an objective way to measure itch is really important across a wide range of medical conditions,” says Shuai (Steve) Xu, a board-certified dermatologist and biomedical engineer at Northwestern University. Itch is a key symptom of not only eczema and other skin disorders, but also liver disease, kidney disease and certain cancers.

In fact, itchiness is one of the main symptoms that reduce quality of life in these patients, says Xu, who co-developed the device with John Rogers at Northwestern“Itch is miserable,” he says. “If you’ve ever had a bout of poison ivy, just imagine living with that day in and day out without reprieve.” 

Measuring and quantifying itch could improve medical treatment for these patients. For example, it would give doctors and patients a more objective way to determine whether a medication is working. That’s particularly true for young children, who may not be able to articulate how their body changes over time. 

Such a device could also improve clinical trials by giving drug companies and regulators objective data that show whether a drug works or not. “There are a lot of new drugs being developed for eczema by drug companies. What do they need? They need objective endpoints to show that their drug works, and the most important one is the quantification of itch,” Xu says. 

Researchers have tried to use standard accelerometers paired with machine learning to monitor scratching, and Nestle Skin Health and Maruho even created an app for Apple Watch called Itch Tracker. But these devices have a tough time capturing scratching when only the fingers are moving, without wrist motion, says Xu. 

To improve upon existing devices, the Northwestern team developed a soft, flexible sensor that captures both motion and high frequency (1,600 Hz) acousto-mechanical signals. These signals are generated by the motion of the hand and the vibrations caused by the fingertips touching the skin. “It allows us to capture the sound of scratching without a microphone,” says Xu. 

The Northwestern researchers initially reported the sensor over a year ago after conducting a study that placed the sensor on the chest to capture signals of swallowing, respiration, and cardiac activity. It was successful, and as a dermatologist, Xu naturally wanted to apply it to skin scratching.  

To adapt the device to scratch, the team trained a learning algorithm—a random forest classifier—to correctly identify scratching. The training involved ten adults scratching a wide range of body locations with the sensor on their dominant hands. “We did a lot to understand what frequency bands a scratch generates, and how it differs from waving and typing and texting,” says Xu.

The image shows two ADAM sensors measuring scratching and sleep quality in a child with eczema.Two ADAM sensors measure scratching and sleep quality in a child with eczema.Image: Jan-Kai Chang

The team then tested the sensor’s accuracy on 11 children with eczema. The children wore the sensor on the backs of their hands while they slept at home for up to three weeks. Parents set up an infrared video camera next to the bed and recorded their children overnight.  

The team then visually reviewed nearly 400 hours of video footage of these kids sleeping and compared it up with time synchronized data from the sensor to see how well the algorithm worked. The data indicated a sensitivity of 84.3% and an accuracy of 99.3%, according to their report. 

Xu says he’s working on a follow-on paper that uses two sensors—one on the hand and one on the chest—to measure both sleep and scratching. He hopes the study will help him better understand the effect of scratching on deep sleep. 

Xu says Novartis and Pfizer financially supported Northwestern’s work on the sensor, and that his spinout company, Sibel Health, is “working with several pharmaceutical companies to implement the sensor as part of their clinical trials.” He said he could not disclose which companies are using it.  

The Conversation (0)