We feel like cough tracking is for anyone with lungs, said Joe Brew, co-founder and CEO of Hyfe AI, the company behind the artificial intelligence app CoughTracker.
But that actually understates the life-changing potential of this little-known but fascinating health-tracking technology.
To illustrate how cough tracking can make a difference and one day even save your life, Brew took us through the interesting beginnings of the CoughTracker app, how AI has grown to understand the difference between a cough and a barking dog, and how mobile technology has evolved. made it possible to monitor the cough in detail.
Why you should monitor your cough
If you’re like me, you probably don’t think much about your cough, but Brew explained why that should change.
Even if you don’t have a chronic cough, you don’t have tuberculosis, you don’t have one of those classic cough diseases, but you get a cough a few times a year just like other healthy people, and that’s valuable information. Just as it’s a useful thing to have a thermometer in your bathroom cabinet, you’re now able to measure this very widespread and very common symptom, Brew said.
By the way, in the future when it comes to treating coughs, cough is the most common symptom that people in doctor’s offices will look back at and ask: How did we do it before automatic cough counting? [Now] they just ask the patient and expect the patient to know, but does anyone actually know how much they cough? Definitely not.
Doctors have been dealing with coughing patients for centuries, so they certainly know everything there is to know about coughs. Given all that, does tracking matter? This is where using artificial intelligence and mobile technology, the same way we count our steps, can change the way doctors deal with them.
One hundred percent of doctors will say, I use cough all the time. I look at coughs, treat coughs, diagnose based on coughs, Brew said. They even use the sound of a cough, so doctors totally understand that it matters. and they know that more cough is bad, less cough is good. But once you get into the super granular stuff, doctors don’t know what to do because they’ve never sat by your bedside all night. There is no body of evidence because when you go to see a doctor, they are with you for 10 minutes and they cannot create a body of data that is useful. They now offered a higher level of accuracy.
Mobile technology makes this possible
Hyfes AI creates and analyzes this data set, but it couldn’t do it without having access to the sound of us coughing in the first place. Brew laughed as he said: If you tried to build a universal cough tracker in 2000, good luck, right? But now we have devices that have all the sensors needed to make cough tracking a reality. Brew sees wearables as the best devices for the job.
I think the ultimate form factor will be the smartwatch, and you can make the app light enough to run on it. The best way [to track coughs] will be built with the watch, and the main reason is precisely its unlimited proximity to the mouth. A lot of things happen to phones that shouldn’t happen if you want to do audio monitoring. They go in purses, they go in back pockets and they stay on the kitchen counter. Having a watch on your wrist is consistent.
The current Hyfes app runs on iOS and Android phones, where the impact on CPU and battery life is relatively low. However, its sound quality gets complicated.
Really cheap phones do a lot of weird things with sound, so our algorithms figured out that this is what a cough sounds like on an iPhone, and this is what a cough sounds like on an $80 phone.
Hyfe lets people comment on the collected audio, which by the way is only collected and analyzed if you agree to do otherwise, everything stays on your device, eliminating privacy concerns. Brew said the human annotation process is necessary for his algorithm to learn because not everyone coughs the same way. He also needs to distinguish between similar sounds so his tracker coughs and not the dog sneezes or barks.
He looks for sharp, explosive sounds. Because of course the cough will not go away or go away. It starts with a loud pop. When it finds an explosive sound, it takes half a sample and sends it to a classifier trained on millions of human-validated sounds to judge whether it was a cough or not.
To get to this point, Hyfe uses a method called a convolutional neural network, where the recorded sound is converted into a spectrogram and, similar to AI recognizing faces in photos, it looks for explosive sounds and then labels them as coughs, dog barks, sneezes or something else, because everyone looks different when examined at this level.
That’s our method, Brew said, adding, and our classifier is right 99% of the time, or something like that absurdly high.
How did this all come about?
We have centuries of audio data to process and we’re always behind, Brew said. We will never catch up with our database in terms of how much audio we captured and how many human annotations there were.
Brew and his team have been working on this since early 2020, when the idea for the cough tracking system came at the start of the COVID-19 pandemic. The first versions of the database were built on data collected from friends and family and personally annotated by the team. Brew admits the first version of the CoughTracker app wasn’t very good because it was hastily put together and launched during the lockdown, but people continued to use it.
The reason they stick around, Brew explained, is because they have a huge problem, which is a chronic cough. It’s not that they have an underlying disease that causes the cough, the cough is the problem. They are hypersensitive and can cough all day and all night, which disrupts their sleep and social life and causes severe irritation. Reality smacked us in the face because this was a real group of people who are totally inadequate.
Brew said some estimates put the number of adults with chronic cough at 30% of the population and said there were not many treatments available because there was no way to measure effectiveness before. A constantly listening cough tracker has the potential to change that, giving people with chronic coughs access to detailed information that doesn’t rely on memory or guesswork. Thanks to this, they can better understand how eating certain types of food affects them, how humidity and air quality change a cough or whether it is an allergic reaction.
Just like when you struggle with blood pressure, you get a blood pressure monitor. If you’re struggling with a cough, having a cough counter in your arsenal really makes sense, Brew said.
Hyphes AI and the importance of tracking and understanding coughs go beyond helping people with chronic coughs. Brew talked about a study that took place in Canada where cough trackers placed near sick patients hospitalized with COVID-19 revealed a reduction in coughing and was an indicator of worsening health rather than improvement.
The really exciting thing is when you go beyond using audio to track coughs, because the same AI and methods can be used to analyze other types of audio as well. Brew talked about using it to help recognize developmental disorders in children, to care for the elderly and ensure regular communication, and to understand other health issues from GERD and sleep apnea to depression.
The future of cough monitoring is bright
You can download and use the Hyfes CoughTracker app now, but what’s next and what’s the distant future for cough tracking?
We think that in early 2024 our wearable cough monitor will be approved [Food and Drug Administration] in the United States, Brew confirmed. That’s what we were hoping for and we’ve been rehearsing right now. But because the regulatory space is so complex and slow and technology moves so fast, we were also really interested in what we could do in the wellness space. which is less about having this particular disease or infection and more about what your trends are and these things can help you improve your life.
As more tech companies focus on health and wellness, cough tracking may become as commonplace as blood oxygen monitoring and respiratory rate monitoring on our wearables. Apple Watch already monitors loud sounds to protect our hearing and uses microphones to detect distinct sounds to enable Crash Detection. Cough tracking using the same hardware doesn’t seem too difficult.
According to Brew, who made this exciting prediction, mobile devices and artificial intelligence will continue to have a positive impact on our health well into the future:
I think that by 2030, our devices will diagnose diseases, continuously monitor and even give you warnings, such as the high probability that you have lung cancer. I think it will definitely come and it will come relatively quickly. They were definitely less than a decade away.
This is an app that originally took two weeks to create before it was launched in the app stores. It now offers centuries worth of data and potentially life-changing (and even life-saving) advice that comes from listening to and analyzing your cough. It’s not hard to see why Brew is so excited about the technology’s potential.
Editor’s recommendation
