FitTech interview: why audio-based AI is a quantum leap for health data

This interview took place after our CEO, Dr Roeland Decorte, was invited to speak as one of 9 speakers for the international FitTech Summit, alongside the Head of Digital Health at Microsoft, the Global Research Leader of Digital Health at IBM, and the VP of Mobile at Adidas. The interview is reproduced here in a written form copied from the FitTech website.

With his company DFI, Dr Roeland Decorte develops Artificial Intelligence that is able to gather the biometric data that the human body is continuously emitting, simply through audio.

Enabling the mass gathering of human body metrics for the first time, the company is exploring a completely new frontier of data.

Before a career developing deep tech, and a PhD at Cambridge in codebreaking, most afternoons and evenings of Roeland Decorte’s childhood were spent in a care home. “My parents founded and operated a private retirement home in Belgium“, he explains. Besides playing card games with the residents, he learned a lot about the human condition, and the factors that impact human longevity. “You would see drastic changes in people, perceive patterns, and begin to realise that no one is consistently observing and mapping these patterns, that there are many situations which could have been prevented.“

He talks to FitTechInsider about how his Artificial Intelligence works, why the future of health data collection lies in listening through AI and what the benefits of his product are.

Your company Decorte Future Industries Ltd develops AI for human health analysis through audio. According to your website your mission is “To augment humanity through the Internet of Bodies.” Can you clarify that?

The Internet of Bodies is essentially everything that connects a human body to the cloud, as well as the data that is being sent from body to cloud. In our case our AI enables a simple microphone to gather thousands of datapoints ranging from cardiovascular to respiratory, which are then transmitted and parsed in real time in the cloud, after which we are able to determine lots of different things about your body.

How do you augment humanity through this Internet of Bodies?

Human health data stands central to many industries – from healthcare to government and insurance to finance – and our technology enables gathering such data on a scale and with a consistency that no one has been able to achieve before.
One core application is in healthcare. Western medicine has long focused on the curative – curing something after it escalated enough to present symptoms. The next step, which our technology now for the first time enables, is shifting to the predictive and preventative, where we are able to intervene before something happens. Or, if certain patterns are observed, even stop it from happening altogether.

On an individual level, being able to cost-effectively, for the first time, continuously monitor the body’s outputs is the first step towards increasing longevity of life, while for businesses having access to human body data can impact everything from their product offering, efficacy, to organisation and logistics. Consider the demand for written data, and now add to that all the unwritten data our body constantly emits but we never collect.

This isn’t science fiction. It’s already happening with GPS and ECG technologies for athletes in professional sports – where athlete data is swiftly becoming the most valuable commodity. The basic nature of the technologies used to gather that data, however, as well as the very limited nature of the data itself – generally just location and heart – means our technology represents a quantum leap.

So, you transform the human body into a controller for the game of life…

Exactly. The main thing that we are revealing is how biometrics – when gathered continuously and en masse – allow us to create a “digital twin” of our own body. At DFI we seek to build a world where our body can directly interact with and inform the digital world. Imagine having real-time predictions and a health bar or status breakdown in the “game of life”, versus simply playing blind, which is what we all do now.

At FitTech Summit 2021 you mentioned that 90% of your data is audio gathered. You said “We use audio to do everything from respiratory rate to heart rate, heart rate variability, gastrointestinal sounds to even neurological based on your voice.” Why does the future of health data collection lie in listening by artificial intelligence?

AI audio analysis is the future, but also backed up firmly by the past. Listening to the human body, in recent centuries using stethoscopes, has for thousands of years been the primary way in which people have cost-effectively assessed someone's overall health. Today we have lots of medical technologies, including portable technologies such as ECG, PPG, while a lot of really good diagnostic medicine is being done in hospitals using large-scale machinery. But time and time again, due to specific requirements from placement to size and power, all these methods have proved impractical to deploy on a continuous basis in daily life.

What are the advantages of your audio based solution?

With our audio-based algorithms we can assess someone's heart rate and heart rate variability as accurately as if they have a full ECG system attached to them. And then we add to that respiratory rate, gastrointestinal analysis, murmur detection, neurological analysis, and many more metrics. As mentioned, our audio is best thought of as creating a digital twin of the human body, which you can keep mining for data almost indefinitely. And all of this just through a single sensor: a microphone.

The fact that we’re using audio means we do all of this without any requirements such as constant skin contact. We don't need apparel to be tight, like all ECG or PPG-based technologies. We don't need electrodes or gel on the electrode. And we don't need to rely on user sweating, as many technologies based on dry electrodes do. What is more, the above technologies require multiple layers of conductive wires and complex compositions to gather data. We just need a microphone. That's it.

So it's a lot more cost effective, it's a lot more intuitive, much more comfortable for the person being monitored. And microphones are already everywhere around us.

Who is your primary target group?

Anyone who has a body, or any organisation that is impacted by human health, is technically a potential client. We are already working with a number of UK and US government agencies and software giants who want to deploy the underlying algorithms in their own technologies or for specific health-related purposes.

I want to move on to the business aspect of the product. I read on Business Weekly in March 2021 that you raised over half a million dollars from US and UK investors. Is this the latest status of your fundraising?

We are currently raising a new $2.7 million seed round, which we want to divide into two tranches, first raising $1M in the next few months, and the another $1.7M by the end of the year.

What is the business model you actually work with?

We provide paid access to our algorithms to enable third parties, from healthcare to government to business, to gather and analyse human body data on an unprecedented scale. We can do this using existing infrastructures – such as third party microphones and devices – or through custom ones. We charge a fixed rate for the use of the SaaS and on the basis of the volume and type of data passing through our systems (based on the biometrics they want to extract).

Which next feature of your product are you currently working on?

The nature of our Artificial Intelligence is that it is constantly expanding, meaning our heart metrics collected through audio are now not just equally accurate as ECG data, but we have started to be able to actively identify and classify heart murmurs.
We want to do further cardiovascular analysis, which is why we're currently training our models to detect specific heart conditions, specifically Atrial Fibrillation, through mass testing and data gathering this year. We are also already able to do voice analysis, which we can use for thing such as mood detection, and cough analysis, which can indicate specific symptoms.

In the near future we will expand into mental and gastrointestinal analysis, and many other metrics you never though were possible through sound, all just through audio. It sounds completely insane, but we for example already know that monitoring audio can be more accurate for assessing when a person has low blood sugar than actually monitoring blood sugar levels themselves. This is because there is a delay, after eating, before effects shows up in the bloodstream. Audio, however, picks up exactly when you eat.

What assumptions about your potential customers did you have at the beginning of your journey?

For a long time we were so determined to deliver value as swiftly as possible that, in testing our tech, we focused on building our own end-to-end solutions and hardware.

This was hugely valuable in increasing our understanding of and maturing the technology, but it was only after conversations with a number of clients, government agencies and tech giants that we realised that the most powerful way to deploy our technology was to allow clients access to our algorithms without our having to necessarily produce the hardware used to collect the audio, or even for us to select the individual use cases. Once we unshackled our technology in this way, the client response was astounding. In short: rather than producing a ready-made end-to-end solution for an industry your client will know better than you, provide your customers with a tool they can deploy in the way they want.

Besides yourself, who do you consider the most influential player in your market?

We are a data company at our core. In general, there are companies out there that have shown how the cycle of gathering lots and lots of data, using it to train your AI to be superior, which invites more clients and thus more data which again makes your AI stronger, works. If you are able to be one of the first to kick off this cycle in an industry, you are holding the keys to the land. When it comes to health monitoring specifically, there's not a lot of companies in the field that are doing continuous monitoring. Direct competitors are mostly ECG-, and PPG-based wearables and intelligent clothing. And, by definition, and for the reasons I mentioned earlier, I don’t think they actually have a solution for the problem they are trying to address. Otherwise, we wouldn't be doing this.

As audio-based health monitoring is a rapidly emerging field, however, there is an increasing number of companies out there that started making applications in our field, though always focusing on just gathering one or two metrics to solve one problem using one device - the “games” to our “game engine” - such as ResApp.

Where is your target market heading?

Governments, businesses, individuals: all are starting to pay a lot of attention to human data. This is being used on a micro level to influence the decisions of companies in, for example, how to deal with or care for individuals, and on a macro level by governments and businesses in making or deploying policies. Human data is the last frontier in the data revolution, and the need for it is continuously expanding.

Populations are aging, many individuals are working remote; we have, for example, an acute crisis coming in the care sector where we won’t be able to accommodate all of us ageing, and there’s at present no solution set out for that.

What about insurance companies? Customers who are monitored by your technology and live a healthy life…

… could get a discount on their insurance. Honestly, I do think that that is where society is headed. It will need to be with the consent of both sides, of course, which is going to be very important. But I definitely see that in the future.