What if you could gauge silent heart conditions like heart failure and irregular heartbeats much before the symptoms showed up or doctors could make sense of them? “AI-enabled screening and monitoring tools could do that, change the way of cardiac diagnosis and prevent a lot of serious conditions,” says Dr Paul Friedman, Chair, Department of Cardiovascular Medicine at Mayo Clinic, USA.
He was in Delhi recently to talk about how AI tools can be helpful in early diagnosis and even work in resource-scarce settings. Excerpts:
How have you used technology at Mayo Clinic to detect underlying heart conditions earlier?
We have been focussing on common medical tests and using AI for screening of heart disease. Let me give you specific examples. The ECG or electrocardiogram, as you know, is a very common test. Someone lays down, they get electrodes to their chest, and typically a piece of paper would come out showing the electrical signals of the heart. It turns out that if you train a computer with hundreds of thousands or millions of such records, it can crunch them, sense patterns and identify very subtle abnormalities even before a human can in a standard ECG.
We trained AI tools to read ECGs the same way a human does, which allows massive scalability. For example, often when someone is having sensations of palpitations or arrhythmias, we give them a monitor they may wear for days or weeks. Then we see if we can record the heart rhythm during those symptoms.
If we see symptoms and heart rhythm at the same time, then we know which heart rhythm abnormality is likely to cause what symptoms and treat the patient accordingly. There are so many hours of reading required that it would not be humanly possible to process data. AI can do that and draw human attention to the abnormalities. This way we can massively scale up our screening capabilities.
In any country, healthcare resources tend to be concentrated in large urban centres while the larger population is in the rural areas. AI-guided tools help us best utilize healthcare resources, which are stretched globally.
We trained a neural network to read a set of 100,000 ECGs and identify abnormal heart pump function or heart failure. We did this over and over again and each time, the machine learnt from human data, refined them and identified the signal indicating a weak heart pump. This machine can read an ECG and tell us something that normally you would need an ultrasound, a CT scan or an MRI scan for. That way you could go into a primary care clinic in a village or a remote area, get a scan and an assessment of whether you need advanced care or not.
Story continues below this ad
Have you tested this module in a large population base?
Yes, we have. We got patients to use a smartwatch and record their ECG within 30 days of a clinically ordered echocardiogram or ultrasound of the heart, a standard test to measure pump strength. A smartwatch or wearable ECG uses a single electrode sensor to measure the electrical activity of the heart, typically placed on its underside. This kind of ECG can detect atrial fibrillation or AFib, which is a type of abnormal heart rhythm. Smartwatch ECGs allow for continuous monitoring of heart activity and provide more convenient access to heart health data than traditional ECG testing.
We took advantage of the data to see whether we could identify a weak heart pump with AI analysis. The test had an area under the curve of 0.88, meaning it was as good as or slightly better than a treadmill test. In other words, its ability to identify a weak heart pump was very strong.
The AI tool roughly identified just under 90% of cases. That’s not too bad compared to not being diagnosed at all. The other key finding was that new diagnosis of a weak heart pump, a potentially life threatening condition that neither the doctor nor the patient knew about, went up by a third, about a 33% increase. So AI is a powerful tool.
Story continues below this ad
Here’s what else was very interesting. In a separate analysis, we found that much of the AI-enabled diagnosis was done by nurse practitioners. So you can really increase the capabilities of the entire healthcare team by making these tools available.
But what of a remote area where everybody may not have a smartwatch?
I would hope that if you put one watch in a clinic and have everyone come in and put it on, you can very inexpensively screen a lot of people.
Have you seen the preventive effects of AI-enabled screening across a large section of the population?
This has been recently approved by the US FDA and we’re still in the process of rolling it out. We hope to mine more data then.
How does this tool help in detecting arrhythmia?
Story continues below this ad
Arrhythmia refers to an irregular heart rhythm. When the heart’s electrical signals are disrupted, even if the heart muscle is normal, healthy, ready to beat, it won’t function until it gets the proper signal. The most common arrhythmia is atrial fibrillation, where the top chambers of the heart quiver. They beat irregularly. The bottom chambers still beat normally, but because the top chambers are beating irregularly, they don’t circulate blood quite as efficiently. The blood pools and clots, one of which may even travel to the brain causing strokes.
People have no symptoms when atrial fibrillation is present, though some people may feel a flutter and a shortness of breath. They may even have exercise intolerance or a little bit of dizziness. But an AI analysis can be like an early warning system.
Shouldn’t this model be studied across populations?
Studies should be across broad heterogeneous populations for them to be representative and accurate. Since there’s a lot of global mobility, migration and intermixing, it’s all the more important to study effects in mixed populations.
How early should one start looking after their heart?
Start young. You could argue that we should start by teaching children at a young age on the importance of physical activity, eating well and navigating social media. In the US, this communication has been mastered by the muppet show Sesame Street. Dr Ruster teaches children about doing things to keep the heart healthy, such as exercising and eating fruits and vegetables. He also hosts health-themed game shows and helps kids deal with stress.
Story continues below this ad
Instead of life span, I’d like to see an increase in the health span. That is the amount of years we live while we’re healthy, where we’re productive, where we’re enjoying the company of loved ones, family, friends.
