Smart sensor patch detects health symptoms through edge computing

Edge computing on a smartphone has been used to analyze data collected by a multimodal flexible wearable sensor patch and detect arrhythmia, coughs and falls.

Wearable sensors are devices that can be worn on the body and measure the state of the body. They are part of the Internet of Things (IoT) and show great promise for monitoring health. These sensors generate large amounts of data, and that data must be processed to be understood. The field of computing dealing with processing these data on the sensor or a device that the sensor is connected to—rather than at a remote server on the cloud—is called edge computing. Edge computing is a key element in wearable sensor technology.

A research team from Japan, led by Professor Kuniharu Takei at Hokkaido University and Associate Professor Kohei Nakajima at The University of Tokyo, have fabricated a flexible multimodal wearable sensor patch and developed edge computing software that is capable of detecting arrhythmia, coughs and falls in volunteers. The sensor, which uses a smartphone as the edge computing device, was described in a paper published in the journal Device.

“Our goal in this study was to design a multimodal sensor patch that could process and interpret data using edge computing, and detect early stages of disease during daily life,” explains Takei.

The team fabricated sensors that monitor cardiac activity via electrocardiogram (ECG), respiration, skin temperature, and humidity caused by perspiration. After confirming their suitability for long-term use, the sensors were integrated onto a flexible film (sensor patch) that adheres to human skin. The sensor patch also included a Bluetooth module to connect to a smartphone.

The team first tested the capability of the sensor patch to detect physiological changes in 3 volunteers, who wore it on their chests. The sensor patch was used to monitor vital signs in the volunteers under wet-bulb globe temperatures (used to determine likelihood of heat stress) of 22°C and over 29°C. “Although our test group was small, we could observe their vital signs change during time-series monitoring at high temperature. This observation may eventually lead to the identifying symptoms of early-stage heat stress,” Takei explains. 

The team developed a machine learning program to process the recorded data to detect other symptoms such as heart arrhythmia, coughing and falls. “In addition to performing the analysis on a computer,” Nakajima elaborates, “we also designed an edge computing application for smartphones that could perform the same analysis. We achieved prediction accuracy of over 80%.”

“The significant advance of this study is the integration of multimodal flexible sensors, real-time machine learning data analyses, and remote vital monitoring using a smartphone,” Takei concludes. “One drawback of our system is that training could not be carried out on the smartphone, and had to be done on the computer; however, this can be solved by simplifying the data processing.” This study advances the concept of a patched-based, edge-computing system for telemedicine or telediagnosis.