Silent heart attacks can now be detected with AI-powered ECG: Why US researcher’s work is a gamechanger

Q: The device you are working with can predict the likelihood of a person having had a silent heart attack. What is the health impact of this?

We are working with two categories of innovations. One, solutions that work in low acuity settings (non-emergency conditions) where someone comes for a check-up, we collect data and screen them for undiagnosed chronic conditions. The solution to detect silent heart attacks is a good example.

They are not having a heart attack right now, they have had it at some point in the past. And, we can make a good guess from the electrocardiogram. Then we can get them on the right medications to maintain lower blood pressure targets, lower cholesterol targets, different anticoagulant medications that are stronger than just aspirin. This can prevent another heart attack.

Besides heart attack, we can also screen people for valvular heart disease, rheumatic heart disease, different kinds of heart failure.

With just the one lead ECG, the predictive power is not good enough to bypass a confirmatory test but we can still rule these conditions out in 90% to 95% of the people. For the other 5% of people, we do a confirmatory ultrasound.

The second category of solutions we’re working on is about diagnosing emergency conditions. So instead of diagnosing someone who has had a heart attack in the past, we’re building algorithms that can tell you if somebody is having a heart attack at the moment. Or if they have a blood clot in the lungs, or an abnormality in the aorta called aortic dissection. By building algorithms with data inside the hospital and then deploying them on these low-cost devices, we can screen people for not just chronic conditions but increasingly start diagnosing these acute problems and routing patients to where they need to go in real time.

Q: Are there other conditions these low-cost devices can monitor or screen for?

Similar tools can be developed for a bunch of metabolic illnesses like diabetes, heart disease, chronic kidney failure. Our pulse oximeter is an interesting innovation. Other than the oxygen percentage, it also gives a waveform to depict the blood flow in the finger. This can give us a lot of information on how the heart is working. The electrocardiogram tells us a lot about the cardio part of the cardiovascular system, but the pulse oximeter tells us a lot about the vasculature.

Q: What are the challenges when it comes to implementation of such innovations?

First is to ensure that the algorithm is right. So, everything we do, we validate rigorously. We’re setting up studies in the ER (emergency room) where people are already in a safe place. We can collect these data and then see what happens to them within the safety of the ER before we start deploying it outside of the hospital.

The next step is figuring out a distribution channel. If you look at how new technologies come into health systems, they slot into things that already exist. But this is different in the sense that no one’s really doing this yet. In the US, for example, it’s really exciting for hospitals to be able to find people with heart attacks that they can treat because health systems are incentivised to find sick people and take care of them. It also fits into preventative care incentive models that governments and insurers have.

So I’m optimistic that there are a number of people who are properly incentivised to help get these tools out there and realise the value to both patients and society.

Q: Why is India a good partner for such AI-based healthcare projects?

Number one, just the volume of data that you can get and, relative to the US, it is much, much cheaper to acquire data here. The other thing is that there is an enormous amount of human capital and talent here. The third ingredient is the rigorous evaluation that you can do here using all the infrastructure that J-PAL (a global research centre based in MIT) has built over the past decades.

This combination makes it an incredibly great place to build innovations that benefit not just India but the rest of the world.

Who owns the data collected for the project?

All of the work that we’re doing is firmly in the public sector. And so, the data, after being very carefully de-identified under all applicable laws, can be shared with researchers who want to build non-profit tools. All of the things that we’re building are in the public domain so that anybody can use them. We feel very strongly that we’re learning from the patients who graciously agreed to participate in the study and the benefits of that research needs to flow back to them.