Your hearing test comes out fine and the bar still swallows half the words + test

• The test says you are fine, the bar says otherwise
• What breaks first is the wiring, not the cells
• A digital twin of your inner ear
• One sound that pokes every synapse at once
• Out of the lab and into a clinic
• A hearing aid cut for one specific ear
• Why this is a twenty-year-old's problem
• The short version
• Five questions: how much noise do your ears put up with? (test)

⇑The test says you are fine, the bar says otherwise

Almost 60 million people in Europe live with hearing loss, and it gets more common with age. The standard check-up is simple: headphones, a run of pitches, and the question of how quiet a tone you can still catch. Land in the normal range and you walk out with a clean result. Then you sit down in a crowded restaurant and half the table turns to mush. That mismatch has a name, hidden hearing loss, and until recently nobody could measure it.

⇑What breaks first is the wiring, not the cells

Hidden hearing loss is thought to come from damage to the connection between the hair cells and the hearing nerve. Researchers call it cochlear synaptopathy, CS for short.

"In 2009, researchers at Harvard found that the first aspect of hearing loss is not losing hair cells in the inner ear, but rather a decrease in the number of synapses," explains EarDiTech project coordinator Sarah Verhulst, from Ghent University in Belgium. "This means that the hair cells may still be intact, but the signals that they send to the brain are not as clear as they should be."

So the microphone still works. The cable running to the brain has fewer strands. In a quiet room that barely shows. In noise, where the brain needs every scrap of detail to pull one voice out of the pile, it shows a lot.

⇑A digital twin of your inner ear

There is no clinical diagnosis for CS and no treatment. Verhulst went after the first half of that. Before EarDiTech started, her team had built CochSyn, a non-invasive prototype test based on a digital signal processing model of the auditory system. A digital twin, in effect: normal hearing reproduced in silico. Then they broke it on purpose, tweaking the model until it behaved like an impaired ear, and watched what changed.

⇑One sound that pokes every synapse at once

What fell out of the model was one very specific sound.

"This specific sound is very good at driving the available synapses at the same time," says Verhulst. "When we play that sound, EEG responses (recordings of the brain's electrical activity) are very large when there are lots of functional synapses. As people start to lose synapses, the EEG signal also starts to decrease."

Play the sound, read the brain's electrical answer off the scalp, and the size of that answer tracks how many synapses are still doing their job. Four things came out of the work:

• a digital model of the auditory system that can be set to healthy or to impaired,
• a sound stimulus, now patented, built to drive all available synapses at once,
• an EEG measurement that shrinks as synapses disappear,
• individualised speech processing algorithms for hearing devices.

⇑Out of the lab and into a clinic

EarDiTech builds on earlier work funded by the European Research Council through the CochSyn project. The European Innovation Council paid for the next stage, which was the unglamorous but decisive part: moving the test from a lab bench into real clinical settings. Clinical trials produced evidence of patient benefit and showed the test holds up in actual clinics.

"These medical trials enabled us to gather results and build a strong business case," she notes.

Results are published, the sound stimulus is patented, and the next step is a spin-off company to put the diagnostic on the market.

⇑A hearing aid cut for one specific ear

Diagnosis was only half the plan. "The aim here was to not only diagnose, but to actively help people who have hidden hearing loss to hear better," adds Verhulst.

Digital models standing in for patients were set to process speech using advanced machine learning. Out came unique individualised audio processing algorithms: not one generic setting, but processing shaped around one person's pattern of damage, meant for personalised next-generation hearing devices.

"We are currently having conversations with partners in the hearing aid and audio chip industry," says Verhulst. "Hopefully we can work towards a licensing strategy to get this innovation to market."

⇑Why this is a twenty-year-old's problem

Because the symptom turns up exactly where the evenings happen. Nobody notices it in a silent room. You notice it in the bar, at the afterparty, in a seminar where six people talk over each other, in the places where you actually want to hold a conversation. Right now that complaint usually ends with a normal audiogram and a shrug. A measurement that can see synapses turns the appointment from "everything looks fine" into "we know what to look for".

⇑The short version

Hearing gives out quieter than the standard test can hear. Ghent built a digital copy of the ear, used it to find a sound that makes the surviving synapses show up on an EEG, ran it through clinical trials on EIC money and patented it. A spin-off is next, and the same models are already writing personalised algorithms for hearing aids. None of it is on a shelf yet, but "I hear fine, I just lose every word in a bar" stops being something you have to explain away.

⇑Five questions: how much noise do your ears put up with? (test)