Using AI to Predict Heart Attacks and Strokes

Artificial intelligence (AI) has been proving its capabilities in the medical world by helping scientists find ways to predict the risks of heart attacks and strokes. In November 2022, researchers at Massachusetts General Hospital as well as Brigham and Women's Hospital's AI in Medicine program presented their findings to the Radiological Society of North America. While the research hasn't yet been published in a medical journal, medical professionals are already buzzing about its possibilities.

What the Study Looked At

Lead researcher Dr. Jakob Weiss and his team were able to predict the risk of death from a stroke or heart attack across a 10-year period by using AI to analyze a single chest X-ray. 

The scientists developed an AI deep learning algorithm, which they labeled the CXR-CVD risk model, and had it analyze more than 147,400 chest X-rays collected from patients who were part of a National Cancer Institute screening trial for prostate, lung, colorectal and ovarian cancer.

Next, the researchers retested the algorithm by having it analyze routine chest X-rays from more than 11,400 outpatients at Mass General Brigham, all of whom were potentially eligible for statin therapy to reduce cholesterol and protect against the possibility of a heart attack or stroke. Over the 10.3 years of the study, 9.6% of those patients experienced a heart attack, stroke or another significant cardiac event. 

After crunching all that data, the AI algorithm was able to meet the clinical standard for predicting future major cardiovascular events such as heart attack and stroke, using just the information found in that single chest X-ray. Per American Heart Association and American College of Cardiology guidelines, doctors currently use a risk calculator that considers variables including age, blood pressure, diabetes, lipid measurements, sex and smoking. Because these variables aren't always known to the doctors, the practical use of the AI algorithm could expand screening significantly, making it possible to assess the risk of cardiovascular disease with just a single chest X-ray.

How AI Predicts the Risk of Heart Attack

The field of radiology has used computer-aided diagnostics (CAD) for years to detect lung nodules on CT scans or breast cancer from mammograms. Today's AI capabilities extend the possibilities of CAD further, thanks to AI's ability to handle huge data sets. Deep learning algorithms can help AI to understand the latent features inherent in imaging data without the need of programmers to create code to do the work.

Doctors typically use chest X-rays to examine the lungs. However, chest X-rays provide a far greater wealth of information, including about the heart and other organs within the field of view of the X-ray. Radiologists — and now AI — can also examine fluid buildup that may indicate heart failure, calcium deposits in the aorta, issues regarding the size and shape of the heart and far more. This information is what the AI involved in the study used to create its predictive model.

What's Next in AI Heart and Stroke Research

Next up for the researchers are further long-term studies, including a controlled, randomized trial. These studies could consider outcomes for patients who have different levels of risk for cardiovascular disease. As the AI deep learning algorithm becomes an assessment tool, it could help identify patients whose risk might otherwise go unnoticed.

Creating a stratification for risk could help doctors assess patients' likelihood of stroke or heart attack up to 20 years in the future, allowing them to take preventative measures. The assessment alone won't prevent heart attack and stroke, but it can open the door for patients and doctors to consider risk levels and choose statins or other treatments to mitigate that risk.