Study shows AI-backed headset system 82.6% accurate at detecting common ear infections, ruptured eardrums and earwax blockage
BUFFALO, NY – New research led by the University at Buffalo shows how headphones could soon detect common ear infections and other ailments.
A study published in June by the Association for Computing Machinery (ACM) describes what the research team calls EarHealth.
The system combines Bluetooth headphones with a smartphone equipped with a deep learning platform. (Deep learning is a type of machine learning, which is itself a form of artificial intelligence.)
EarHealth works by sending a whistle through a healthy user’s headphones. It records how the chip reverberates in the ear canals, creating a profile of the unique geometry of each user’s inner ear.
Subsequent whistles (for example, a user can set the system for a daily test) monitors each ear for three conditions that alter the geometry of the ear canal. These conditions are earwax blockage, ruptured eardrum, and otitis media, a common ear infection.
Each condition has a unique audio signature that the deep learning system can detect with fairly accurate results.
The researchers reported that EarHealth achieved 82.6% accuracy in 92 users, including 27 healthy subjects, 22 patients with a ruptured eardrum, 25 patients with otitis media, and 18 patients with earwax blockage.
“With people around the world living longer and the prevalence of headphones, it’s more important than ever to monitor your hearing health,” says lead author Zhanpeng Jin, PhD, associate professor in the Department of Computer Science and Engineering from the University at Buffalo. School of Engineering and Applied Sciences.
“With EarHealth, we’ve developed what we believe is the first headphone-based system that monitors ear health conditions in an effective, affordable and easy-to-use way,” he adds. “Because it has the potential to detect these conditions very early, it could greatly improve health outcomes for many people.”
The study was supported by the US National Science Foundation and was presented in June at the ACM International Conference on Mobile Systems, Applications and Services (MobiSys) in Portland, Oregon.
Yincheng Jin, a PhD candidate at the University at Buffalo, is the first author of the study.
Co-authors include Yang Gao, PhD, a postdoctoral researcher at Northwestern University, and Zhengxiong Li, PhD, assistant professor of computer science and engineering at the University of Colorado Denver. Both received their doctorate from the UB.
Other co-authors include Xiaotao Guo, PhD, a researcher at The First Affiliated Hospital of USTC in China, and Jun Wen, PhD, a postdoctoral fellow at Harvard Medical School.
The team is planning further studies to refine the system. These include testing how ear hair, a history of eardrum inflammation and other factors can affect EarHealth’s performance.