The booth is the cost
ยท 7 min read
A hearing test is one of the cheapest diagnostic procedures in medicine. You play a tone, the patient raises a hand, you write down the quietest level they heard. The information content is tiny. A trained person can run a screening in ten minutes.
So why does almost nobody in rural India get one?
Because the test is cheap and the room is not. To produce a threshold you can put in a chart, you need a sound treated booth, a calibrated clinical audiometer and somebody trained to run both. That bundle is a capital purchase and a permanent floor plan commitment, which means it lives in district hospitals and private clinics in cities. The patient has to travel to the room. Most never do, and by the time they arrive the loss is bad enough that they noticed it themselves, which is exactly the point at which screening stops being useful.
The bottleneck in hearing screening is not the test. It is the room the test has to happen in.
Why the room exists
The booth is not there for comfort. It is there because of arithmetic.
When you measure a hearing threshold you are looking for the quietest sound a person can detect. For normal hearing that can be around zero decibels hearing level, which is a genuinely tiny amount of acoustic energy. If the room has ambient noise at the same frequencies, that noise masks the tone and the threshold you record is the room's, not the patient's.
Standards bodies put numbers on this. There are published maximum permissible ambient noise levels for audiometric test rooms, specified per frequency band and varying by how low you need to test and whether you are using supra-aural headphones, insert earphones or a free field. The booth is simply the cheapest reliable way to hit those numbers in a building that was not designed for it.
So the booth is doing one job: making the noise floor low enough that a quiet tone is still measurable. Once you see it that way, it stops being sacred and starts being one solution to a specific constraint.
The second thing the booth is hiding
There is a subtler reason clinical audiometry is locked to expensive equipment, and it took me a while to appreciate it properly.
Decibels hearing level is not a physical unit. It is defined relative to a reference threshold for normal hearing, and that reference is transducer specific. A given supra-aural headphone and a given insert earphone need different sound pressure levels at the eardrum to produce what the standard calls zero. Those reference values are published per transducer type, and a clinical audiometer is trusted because it has been calibrated against them, with a coupler, on a schedule.
This is the real reason you cannot just play tones through consumer earbuds and call it an audiogram. The output level at any given digital volume is unknown, it varies between units, it varies with fit, and it drifts. You would be producing numbers with the shape of an audiogram and none of the meaning.
What we are actually building
At Neurento the bet is that both requirements can move into the device.
The calibration half is tractable because it is a manufacturing problem rather than a clinical one. If you control the transducer, and you characterise that exact model properly once, then every unit coming off the line inherits a known relationship between digital level and output level. You are not calibrating in the field. You are shipping a device whose output is already known, and verifying rather than establishing.
The ambient noise half is more interesting, because you cannot make a school corridor quiet. What you can do is stop pretending you do not know how loud it is. A microphone on the device can measure the noise floor per frequency band during the test. From there you have honest options. You can raise the lowest testable level and report that you did. You can wait for a quiet interval. You can mark a frequency as untested rather than reporting a threshold you cannot defend.
That last one matters more than it sounds. A screening tool that says I could not test 500 hertz in this room is far more useful than one that quietly returns a bad number. The failure mode of cheap diagnostics is not usually imprecision. It is confident imprecision.
What this is not
It is not a replacement for diagnostic audiometry. A booth and an audiologist produce a clinical audiogram, and there are decisions, hearing aid fitting among them, that should keep depending on that.
What we are going after is the tier below, which currently does not exist for most people. Screening. A yes or no on whether this person needs to see someone, done in a room that already exists, by someone who is already there. If that works, the booth stops being the entry price to the entire field and becomes what you escalate to.
Why I care about this specific problem
I came at hearing from the wrong direction. My screening work started in cognition, and the clinical trial for it runs across two departments, ENT and Neurology, which surprised people when I described it.
It stopped surprising me once I read the literature. Hearing loss and cognitive decline travel together far more often than either field tends to emphasise, and the causal story is still argued over. What is not argued over is that untreated hearing loss is common, cheap to detect in principle, and routinely missed.
A dementia screening programme that ignores hearing is measuring a confounder and calling it a finding. That is what pushed me from software into transducers. I got tired of writing models that inherit whatever the sensor decided to give them.
If you work in audiology, acoustics or device calibration and you think this is naive, I would genuinely like to hear where. isararsiddique@gmail.com