Echo Strain AI – full auto strain analysis using artificial intelligence

Cardiac Wire Q&A with Us2.ai president Yoran Hummel to discuss fully automated strain analysis using Us2.ai’s echocardiography decision support platform. 

Can you share a little bit about yourself and Us2.ai?

My name is Yoran Hummel. Based in the Netherlands, I’m one of three Us2.ai co-founders and the current acting president. I’ve had a long career in echocardiography for approximately 25 years doing echoes myself. Eventually, I got fed up with echo, not so much with the cardio mechanics, or the hemodynamics of echo, or even my guilty pleasure, the physics of ultrasound. I got fed up with the manual, labor-intensive work. Searching for the best images, searching for the best frames, and doing endless drawings of cavities and cardiac structures. And that’s how I ended up being a founder of this company called Us2.ai.

How would you describe Us2.ai to people who aren’t familiar with the company?

Us2.ai is a problem solver that automates the analysis of echo. It brings a tedious, 30-minute procedure back to two minutes, single click, full analysis. And when we say ‘full’, it’s full analysis with no human intervention, yet with the possibility to do everything manually. So we open up the black box approach to analysis.

One of the biggest problems that has always been an issue within echocardiography, is human variability. For example, if I measure an echocardiogram, and I remeasure it later that same day, my results will be different. This is inherent to echo. This is inherent to being a human. We are simply not machines that follow a program. 

Automation, working with Machine Learning, actually gets rid of that variability. It’s objective because it does the same thing, every time. The automated algorithm doesn’t get distracted; it simply does what it’s trained to do. This means that in longitudinal follow-up of patients, you get an objective measure through time, which helps us to track the changes more accurately. This is potentially a whole new spectrum of information from which we can learn more about cardiac disease.

Word on the street is Us2.ai will soon offer strain analysis. What is strain? Why is it important? And how is it different from conventional methods?

Strain is a measure of cardio-mechanics. It measures how the muscle of the heart works. Strain allows us to look at the muscle and at the function of the muscle. The muscle, and its function, actually dictate how the blood flows within the heart. In contrast, in conventional echo, we measure all kinds of blood-related flows, like velocities and volumes. Most of our conclusions are based on those volumetric and flow hemodynamics instead of the thing that actually causes those hemodynamics to change – the cardiac muscle. With strain, we have the opportunity to measure the cause of those hemodynamic changes directly.

How does strain analysis satisfy the needs of patients, doctors, and other stakeholders?

What we call ‘cardiac disease’ is, in most cases, a disease of the muscle. Since we now have the possibility to actually examine the function of that muscle, we are closer to the cause of the disease. 

How does it affect the work of a doctor? We are now able to detect the subtle changes in muscle function caused by disease much earlier. This potentially enables us to intervene earlier. And that’s a huge game changer of course, in cardiovascular disease, which is currently one of the biggest killers globally.

When this launches, how long until physicians will be able to use this feature in their day-to-day practice?

Immediately. Keep in mind, strain itself is already mostly accepted. However, full automation of strain is not. In a normal setting, what you need to do is find an image, then pinpoint certain landmarks on that image to then start the analysis. But the Us2 solution is a non-click solution. As long as a user provides us with the images, strain will be included in your normal reporting. So it’s not an add-on, it’s not an extra, it’s an integral part of your conventional echo.

So there’s no additional work on behalf of the physician or the person performing the echo?

None whatsoever.

How does the addition of the strain analysis fit into the company’s overall strategy?

The aim of our company is to democratize echo. Everyone can echo, that’s our tagline. And to do so, we must aim for reducing known variability in all the analyses. So adding fully automated strain is just the next step for us as a company. To make it the best clinical tool we can, we need to be able to do everything fully automated. Our first release was a good basis of structural and functional measurements. And now we’re adding on all the new things! Strain is just one of many exciting things coming. We’re working hard towards a full-blown echo with everything you could ever imagine. Fully automated!

Why do you think there is hesitancy in the healthcare sector to embrace AI solutions? 

I think the main driver is people simply think it’s this big unknown. Which is weird, because AI is everywhere already. Maybe there’s also a bit of fear factor in it. People sometimes wonder, ‘Is it going to take my job?’ What might help is to take a look at what we’re doing now in healthcare. Is that sustainable? I think most people, deep in their core, know that it isn’t. We need to embrace automation. It’s not man versus machine, it’s man and machine. This tool is not replacing you. It’s supersizing you.