The first telemedicine session ever took place in 1897, when doctors made a remote diagnosis of croup based on the sound of a cough during a telephone consultation.
Now, healthcare company Hyfe uses acoustic AI in its digital cough monitoring.
“Digital cough monitoring is Hyfe’s bread and butter,” says Dr. Joe Brew, Co-Founder and CEO of Hyfe AI. “It consists of processing continuous audio so as to identify cough, and then quantifying cough’s frequency over time.”
Acoustic AI and coughing
As the world’s most prevalent symptom, and the number one reason people go to the doctor, coughing is hugely neglected. “How’s your cough?” is a typical question that a caretaker asks a patient, but there is no clear answer for it.
“Everything is quantified in medicine and wellness - heart rate, weight, blood pressure, etc. - but cough never really has been,” says Brew. “We want to leverage the billions of microphones around the world so as to detect and quantify coughing, improving the lives of anyone with a cough, which is all of us, at least some of the time!”
Hyfe’s flagship model is named ‘Alison’.
“We named it ‘Alison’ because it is ‘ALways LISTENing’,” says Brew. “It takes raw audio from the microphone, detects cough-like explosive sounds, and then categorises each of those sounds as cough or non-cough. When a cough is detected, the timestamp is saved, allowing us to construct the history of exactly when a user coughed. We also develop models which analyse the acoustic characteristics of each cough, and cross that data with other data streams (nearby weather conditions, heart rate, etc.), but our focus is simply counting coughs via sound.”
AI is changing everything, and the pace of change will only increase in the coming years. Much of the innovation has been focused on images and text, but audio offers huge opportunities, especially since so many devices (smartphones, wearables, smart speakers, etc.) have microphones. Sound is extremely information-rich, so it’s a good area to work on for AI in health.
“A lot of innovation never makes it to the end user due to usability. Unlike others working in the cough space, we’re focused on longitudinal, background cough detection and quantification, so that the user doesn’t have to do anything. That is, one can count their coughs the same way they might count their footsteps - passively, in real-time, all the time. This means that changes in cough frequency can be detected prior to the user even being aware.
“In the coming years, we imagine a world where everyone monitors their coughing all the time, using Hyfe’s lightweight, on-device algorithms, being able to track their symptoms when ill, and potentially being able to detect illness onset prior to those symptoms being noticeable to them.”
How can this support the world post-COVID-19
There are a lot of applications for cough tracking in the post-COVID-19 world.
“First, COVID-19 is here to stay, and tracking cough during illness onset and recovery can be extremely useful,” says Brew. “Second, long COVID-19 patients often have coughs. Third, the detection of novel outbreaks can be supported by disease-agnostic syndromic surveillance / symptoms tracking. Fourth, other respiratory pandemics will still be here long after the world stops focusing on COVID-19 (for example, Tuberculosis). Finally, chronic cough is a major issue for millions, and cough tracking can significantly help chronic coughers in the management of their condition.”
By 2024, Brew wants Hyfe AI to get its FDA-cleared cough counting watch (the “Hyfe Cough Monitor”) by early 2024.
“We hope this will bring cough tracking to the many millions of people with chronic cough who could benefit from objective symptoms data,” he says. “We want to continue the growth of our flagship phone products (the “Cough Tracker”, which we’ve built with Merck in the USA; and the “Cough Pro” elsewhere in the world). We want to unveil our first digital therapeutic for the treatment of chronic cough.”
Brew wants to continue to develop a research-driven insights layer so that the science of cough advances, and so that users can understand how and why they cough.