Researchers at Western University are developing a new process for creating hearing aid earmolds for children. The ALLEars project uses artificial intelligence to predict a child's ear growth and 3D printing to fabricate the custom molds. This approach seeks to reduce the number of clinical visits and interruptions in hearing aid use during a child's critical language development years. The project combines expertise in audiology, engineering, and data science to address a persistent challenge for children with hearing loss and their families.
Full transcript
For a parent, the journey can start with an unexpected diagnosis for their three-month-old son: hearing loss.
And with that diagnosis comes a piece of technology—a hearing aid. But for an infant, there's a problem, and it's not the electronics.
It's the earmold. That soft, custom-fit piece that holds everything in place. A baby's ears grow, and they grow fast.
So fast that the earmold stops fitting. It gets uncomfortable, it falls out, or it creates a high-pitched whistle.
Which means another trip to the clinic. Another session where a toddler has to sit still while silicone is put into their ear to make a cast. And then, weeks of waiting for a new one to be made.
In that gap—that waiting period—a child loses consistent access to sound. This happens over and over, right during the few years when their brain is learning language.
It's a race against a biological clock. So what would it take to change the race? What if, instead of just keeping up, you could get ahead of a child's growth and make an earmold for an ear that doesn't exist yet?
For a parent, learning that their infant has hearing loss is the start of a journey through many unknowns. One of the persistent challenges they face is a small, custom-fitted piece of silicone called an earmold. A hearing aid works by capturing sound, processing it, and sending it through a tube to this earmold, which directs the sound into the ear canal. For the system to function correctly, the earmold must fit perfectly.
But for infants and toddlers, a perfect fit is a temporary state. This is a period of rapid physical growth, and a child’s ears are no exception. According to Susan Scollie, a professor in the Faculty of Health Sciences at Western University, a young child might outgrow their earmolds in as little as three months. When an earmold becomes too small, it can cause discomfort, produce a high-pitched whistling sound known as feedback, or simply fall out. The quality of the sound being delivered to the child’s ear is diminished.
This cycle of outgrowing earmolds occurs during the most critical period for a child's language acquisition and brain development. Consistent auditory input is a foundation for learning to process speech and to speak. Scollie notes that each time a child is waiting for a new, better-fitting earmold, their access to sound is interrupted. This creates recurring gaps in their developmental journey.
The conventional process for getting a new earmold is burdensome. A family must travel to an audiology clinic where a specialist injects a soft silicone putty into the child's ear. The child has to remain still for several minutes while the putty hardens into a perfect cast of their ear. This experience can be stressful and uncomfortable for an infant or small child. Once the impression is made, it is sent away to a lab for fabrication, a process that can take weeks. This entire sequence repeats every few months for the first few years of a child's life.
To address this long-standing clinical problem, a large-scale project at Western University called ALLEars is developing a new, technology-based approach. The project, a collaboration between Western’s Faculty of Health Sciences and Faculty of Engineering, received a 4.4 million dollar grant from the Oberkotter Foundation to re-engineer the entire process.
The first part of the solution uses artificial intelligence to predict the future. Soodeh Nikan, an Assistant Professor in the Faculty of Engineering, is leading the development of a predictive model. The system is trained on a massive dataset, containing thousands of 3D scans of earmolds from children of different ages, biological sexes, and ethnicities. By analyzing this data, the AI model learns the complex, non-linear patterns of how an ear grows and changes shape over time.
This allows the ALLEars team to digitally “grow” a model of a child’s ear. From a single 3D scan, the AI can predict what that child’s ear will look like in several months. The result is that a new earmold can be designed and ready for the child *before* they have outgrown their current one, effectively eliminating the gap in hearing access. The AI also has another function: mirroring. Nikan explains that the model can learn the typical, subtle differences between a person's left and right ears. This would allow a clinician to take an impression of only one ear and use the AI to generate an accurate model for the other. For a child who finds the impression process frightening, cutting the procedure in half would be a significant relief.
Once the AI has generated a digital design for a future earmold, the next step is manufacturing. This is where Joshua Pearce and his team come in. Their goal is to create a method for producing these custom earmolds that is extremely low-cost, fast, and can be done anywhere in the world. The solution is 3D printing. The team is refining every part of the process, from the hardware of the printers to the software that converts the AI’s design into printable instructions.
Pearce notes that traditional manufacturing has many manual steps. The ALLEars approach seeks to remove all of them except for final assembly. This could shrink the production timeline from weeks to a matter of hours. A clinic could potentially print multiple earmolds for multiple patients at the same time, right on site. If a child loses or damages an earmold, a replacement could be printed quickly and at low cost.
Central to this part of the project is a commitment to open science. All the software, firmware, and hardware designs the team develops will be released with an open-source license. This means a clinician in a remote Canadian community or a health worker in a developing nation will have free access to the same tools. The goal is to create a complete, replicable toolchain that allows anyone to leverage the technology, ensuring the benefits are distributed as widely as possible.
The ALLEars project involves multiple collaborators, including a partnership with Boys Town National Research Hospital in the United States, which is contributing data from its extensive pediatric audiology research to help train the AI. As the project progresses, the team plans to recruit other partners globally to ensure their dataset is representative of children from all backgrounds.
Dr. Teresa Caraway, the CEO of the Oberkotter Foundation, characterizes the project as an application of evolving technology to address a persistent challenge in pediatric hearing healthcare. For families, the outcome of this work could mean fewer clinic appointments, less stress, and more security. For a child with hearing loss, it could mean something more fundamental: the promise of continuous, uninterrupted access to the world of sound during the years when it matters most for their development.
So the work of researchers like Susan Scollie, Soodeh Nikan, and Joshua Pearce at the ALLEars project comes down to one objective: closing that gap.
Using new tools to predict how a child’s ear will grow, and then printing the next earmold before it’s even needed. No more waiting, and no more interruptions to hearing.
And the plan is to make the entire process open-source, so a clinic anywhere in the world could have the tools to do this on site.
This research is made possible by a grant from the Oberkotter Foundation and collaboration with partners including Boys Town National Research Hospital.
That’s our show. You’ve been listening to the FHS TestPod.