Photo Credit: iStock.com/TefiM

Using data from wearable devices, an AI algorithm successfully distinguished patients with severe knee osteoarthritis from healthy controls.

A machine learning algorithm that analyzed kinematic data from a wearable device effectively differentiated patients with severe knee osteoarthritis (OA) from healthy controls. Researchers published the findings in the Journal of Medical Signals & Sensors.

“Inertial measurement unit (IMU), comprising accelerometer, gyroscope, and magnetometer, is a rapid, cost-effective, and noninvasive method for collecting kinematic data,” wrote corresponding author Saeed Kermani, PhD, and colleagues. “Considering the side effects of repeated exposure to ionizing radiation, X-ray device wear and tear, and lack of access to radiography facilities in remote areas, there is an increasing need for primary automated OA diagnosis by sensor variable.”

Several published studies have investigated machine learning models that use IMU data for diagnosing OA and tracking progression of the disease. This study evaluated the feasibility of a wearable device for distinguishing between 15 healthy individuals and 15 patients with grade 4 knee OA.

“The wearable device consisted of 2 IMU sensors, one on the lower leg and one on the thigh,” researchers explained. “One of the sensors is used as a dynamic coordinate system to improve the accuracy of the measurements.”

Participants were instructed to take eight typical steps in 10 seconds without the use of any assistive tools.

A total of 1,433 IMU signals were collected from participants, and various machine learning models analyzed the data. The best performing model achieved an accuracy of 93.71 ±1.1 and a precision of 93 ±1.3 in differentiating patients with severe knee OA from healthy controls, according to the study.

“The proposed algorithm outperforms existing methods in similar articles in sensitivity,” the researchers wrote, “showing an improvement of 4%.”

Because the main objective of the study was to investigate the feasibility of a wearable device for diagnosing knee OA, the extreme comparison of healthy individuals with patients with severe knee OA may have affected the strength of the reported results, researchers noted.

“Therefore, it is possible that in a study on milder forms of KOA [knee OA], the results obtained with the mentioned method may decrease,” they wrote.

Future studies will include patients with grades 2 and 3 knee OA.