An Earful of AI

By age 70, about two-thirds of us will experience age-related hearing loss. That’s the bad news. The good news: Hearing aid technology is improving all the time, in large part thanks to artificial intelligence (AI) incorporated into devices.

Shen Yi, associate professor of speech and hearing sciences, focuses on the potential of AI to provide increasingly individualized solutions for hearing aid users. He heads the Applied Hearing Science Lab in the UW Department of Speech and Hearing Sciences.

“As hearing aids and cochlear implants are getting more complex, the devices have so many parameters you can tweak,” says Shen. “Ideally, you want to set those parameters to meet the patient’s unique needs. Individualization is the key thing I’m interested in.”

Shen describes the mechanics of hearing as “somewhat of a miracle.” Consider all the things that must happen for a person to detect sounds and assign them meaning, from a child crying to a water faucet dripping to a friend telling a joke.

First, sounds create pressure waves that enter the ear and push on the eardrum and the tiny bones of the middle ear. Then those pressure waves transition to fluid motion inside the cochlea, or inner ear. Sensory hair cells pick up that fluid motion and send information to neurons in the brain’s auditory system. That information is translated into a neural code that the brain can understand.

A lot can go wrong along the way. Genetic mutations can cause parts of the ear to malfunction. Ear infections can affect the middle ear, though that condition is usually temporary. The most common problem is damage to the hair cells in the cochlea, either due to noise exposure or aging. Hearing aids work to bolster the damaged hair cells by amplifying sounds.

“The challenge of hearing aids for the longest time has been that they amplify the things you want to hear, but also things you don’t want to hear,” says Shen. “They are less selective than you would want them to be. That’s where the newer generation of devices with AI in them makes such a difference.”

Shen describes AI in hearing aids as “kind of a little brain” that amplifies what you want to hear and eliminates background noise before it reaches your ear. Algorithms for making this happen have been around for a while, but with more efficient algorithms, the technology can now be implemented in mobile and wearable devices without severely impacting battery life of the devices.

An area of improvement that Shen is hoping to address is the ability of hearing aid users to “train” their device to respond to their specific needs without having to fuss with complicated technology or understand the underlying AI. With support from the National Institutes of Health, his team is developing an app that generates feedback on sound quality using an interactive approach. As the user answers a series of multiple-choice questions to compare how several hearing aid settings perform in the user’s current environment, the AI model fine-tunes the hearing aid to achieve optimal sound quality.

The interface is simple for the user, but in the background, a complex computational infrastructure is making adjustments and optimizing the options presented to the user in response to the user’s previous choices. Research participants — older adults who are hearing aid users — are currently testing the app.

The multiple-choice approach may bring to mind eye exams where the patient compares two lenses, then another two, and so on, to identify the appropriate prescription. “It’s a similar process,” Shen says, “except you need to imagine that you buy eyeglasses from an online store, receive them, and connect them to an app on your phone where you play some kind of interactive game for five minutes, after which the lenses change to your optimal prescription for your needs in that moment. That’s the kind of technology we’re talking about.”

Shen's goal is to empower people with hearing devices so they have more control on their own. “That’s what these kinds of technology are designed for, and I think they are on a rising trajectory," he says. "I believe they will meet people’s individual needs better and better.”

Shen’s team is exploring a somewhat similar approach with cochlear implants (CI), a hearing device for people with severe to profound hearing loss in one or both ears. But the complexity of CI technology presents greater challenges. Unlike hearing aids, which work with a user’s existing hearing, cochlear implants place electrodes in the cochlea to stimulate neurons in the brain — the job sensory cells are no longer able to do.

When people receive cochlear implants, it takes time and training for their brains to understand and adapt to the electronic sounds produced by the implant. When a patient says they don’t like the sound quality, it can be difficult for a clinician to know whether it’s the usual frustration of a new CI user or an actual problem with the device. 

“Users often cannot describe what’s wrong with enough specificity for their audiologist to identify the problem,” says Shen. To address this, Shen is developing an app that systematically collects feedback from CI users to help pinpoint the issue. Results from the test would provide the audiologist with better information to make adjustments to the device. His team is partnering with UW Otolaryngology Head and Neck Surgery for this research.

The team also partners with Seattle Children’s hospital, for a project that uses AI to assist with hearing diagnosis in babies and toddlers. Since young children cannot communicate what they are experiencing, hearing tests involve placing electrodes on a child’s head, presenting a sound, and reviewing the brainwaves to analyze the response. The test is time-consuming and requires an expert audiologist.

Many places do not have audiologists with this expertise, so Shen’s team has been developing an AI algorithm to enable troubleshooting and analysis of test results in real time, just like an expert audiologist would. This technology can also be helpful in training new audiologists.

Shen is collaborating with Intelligent Hearing Systems, a major manufacturer of testing equipment, with the goal of building this technology for testing into their devices and putting it into clinics.

“It’s an exciting time in the field of audiology,” says Shen. “As the technology for hearing testing and hearing aids continues to improve with advancements in AI, many millions of people experiencing hearing loss will benefit. I’m glad to be part of this effort.”