1st smartphone app can detect ear infections in children
- School of Medicine: Brian Donohue - 206.543.7856, firstname.lastname@example.org
- School of Computer Science & Engineering: Sarah McQuate - 206.543.2580, email@example.com
Ear infections are the most common reason that parents bring their children to a pediatrician, according to the National Institutes of Health.
This condition occurs when fluid builds up in the middle ear behind the eardrum and is infected. This buildup is also common in another condition called otitis media with effusion. Any fluid buildup can be painful and make it hard for children to hear, which can be especially detrimental when they are learning to talk.
Both conditions are hard to diagnose because symptoms are vague: Children might tug on their ears or have fevers, or be asymtomatic. Young children may be unable to describe where they hurt.
University of Washington researchers have created software that can detect fluid behind the eardrum by simply using a piece of paper and a smartphone's microphone and speaker. The smartphone makes a series of soft audible chirps into the ear through a small paper funnel and, depending on how the chirps are reflected back to the phone, the app determines the presence of fluid with a detection probability of 85%. This is on par with the best current diagnostics, which involve instruments using acoustics or puffs of air.
The team published its results May 15 in Science Translational Medicine.
"Designing an accurate screening tool on something as ubiquitous as a smartphone can be game-changing for parents as well as healthcare providers in resource limited regions," said co-author Shyam Gollakota, a UW associate professor at the Paul G. Allen School of Computer Science & Engineering. "A key advantage of our technology is that it does not require any additional hardware other than a piece of paper and a software app running on the smartphone."
A quick screening at home could help parents decide whether or not they need to take their child to the doctor. Once diagnosed, ear infections can be easily treated with observation or antibiotics. Persistent fluid can be monitored or drained by a doctor to relieve symptoms of pain or hearing loss.
This app works by sending sounds into the ear and measuring how those sound waves change as they bounce off the eardrum. The team's system involves a smartphone and ordinary piece of paper that a doctor or parent can cut and fold into a funnel. The funnel rests on the outer ear and guides sound waves in and out of the ear canal. When the phone plays a continuous 150 millisecond sound — which sounds like a bird chirping — through the funnel, the sound waves bounce off the eardrum, travel back through the funnel and are picked up by the smartphone's microphone along with the original chirps.
"It's like tapping a wine glass," said co-first author Justin Chan, a doctoral student in the Allen School. "Depending on how much liquid is in it, you get different sounds. Using machine learning on these sounds, we can detect the presence of liquid."
When no fluid is present, the eardrum vibrates and sends a variety of sound waves back. These sound waves interfere mildly with the chirp, creating a broad, shallow dip in the overall signal. When fluid is present, the eardrum doesn't vibrate as well and the sound waves, reflected back, interfere more strongly with the chirp and create a narrow, deep dip in the signal.
To train an algorithm that detects changes in the signal and classifies ears as having fluid or not, the team tested 53 children between ages 18 months and 17 years at Seattle Children's Hospital. About half of the children were scheduled to undergo surgery for ear tube placement, a common surgery to relieve chronic or recurrent bouts of ear fluid. The other half were scheduled to undergo a different surgery unrelated to ears, such as a tonsillectomy.
"What is really unique about this study is that we used the gold standard for diagnosing ear infections," said co-first author Dr. Sharat Raju, a surgical resident in otolaryngology-head and neck surgery at the UW School of Medicine. "When we put in ear tubes, we make an incision into the eardrum and drain any fluid present. That's the best way to tell if there is fluid behind the eardrum. So these surgeries created the ideal setting for this study."
After parents provided informed consent, the team recorded the chirps and their resulting sound waves from the patients' ears immediately before surgery. Many of the children responded to the chirps by smiling or laughing.
Among the children getting ear tubes placed, surgery revealed that 24 ears had fluid behind the eardrum, while 24 ears did not. For children scheduled for other surgeries, two ears had bulging eardrums characteristic of an ear infection, and the other 48 ears were fine. The algorithm correctly identified the likelihood of fluid 85% of the time.
Then the team tested the algorithm on 15 ears belonging to younger children between 9 and 18 months old. It correctly classified all five ears that were positive for fluid and nine out of the 10 ears, or 90%, that did not have fluid.
"Even though our algorithm was trained on older kids, it still works well for this age group," said co-author Dr. Randall Bly, a UW assistant professor of otolaryngology-head and neck surgery at the School of Medicine and Seattle Children's practitioner. "This is critical because this group has a high incidence of ear infections."
Because the researchers want parents to be able to use this technology at home, the team trained parents how to use the system on their own children. Parents and doctors folded paper funnels, tested 25 ears and compared the results. Both parents and doctors successfully detected the six fluid-filled ears. Parents and doctors also agreed on 18 out of the 19 ears with no fluid. In addition, the sound wave curves generated by both parent and doctor tests looked similar.
"The ability to know how often and for how long fluid has been present could help us make the best management decisions with patients and parents," Bly said. "It also could help primary care providers know when to refer to a specialist."
The team also tested the algorithm on a variety of smartphones and used different types of paper to make the funnel. The results were consistent regardless of phone or paper type. The researchers plan on commercializing this technology through a spinout company, Edus Health, and then making the app available to the public.
"Fluid behind the eardrum is so common in children that there's a direct need for an accessible and accurate screening tool that can be used at home or clinical settings," Raju said. "If parents could use a piece of hardware they already have to do a quick physical exam that can say 'Your child most likely doesn't have ear fluid' or 'Your child likely has ear fluid, you should make an appointment with your pediatrician,' that would be huge."
Rajalakshmi Nandakumar, a doctoral student in the Allen School, is also a co-author on this paper. This research was funded by the National Science Foundation, the National Institutes of Health (5T32DC000018-35), and the Seattle Children's Sie-Hatsukami Research Endowment.
For details about UW Medicine, please visit http://uwmedicine.org/about.