Artificial Intelligence

AI Can Predict Heart Attack or Stroke Death Risk From an X-Ray

New AI deep learning algorithm uses a single X-ray to predict health risks.

Posted December 5, 2022 | Reviewed by Davia Sills

Key points

Thanks to artificial intelligence (AI), researchers are advancing ways to detect health risks early.
A deep learning AI model can predict someone's 10-year risk of death from a stroke or a heart attack from a single chest X-ray.
This could enable patients to make healthy lifestyle changes to reduce the danger.

Source: Geralt/Pixabay

The pattern-recognition capabilities of artificial intelligence (AI) deep learning are rapidly advancing potential new ways to detect health risks early. A new study by researchers at Massachusetts General Hospital and the AI in Medicine program at the Brigham and Women’s Hospital in Boston demonstrates how an AI deep learning model can predict the 10-year risk of death from stroke or a heart attack from a single chest X-ray. The researchers presented their findings at the recent Radiological Society of North America (RSNA) 2022 annual meeting.

Cardiovascular diseases (CVDs) are the leading cause of death worldwide, with an estimated 15.2 million deaths globally due to heart attack and stroke in 2019, according to statistics from the World Health Organization (WHO). Early detection of cardiovascular disease gives the opportunity for potentially life-saving interventions and treatments. Most cardiovascular diseases can be prevented by changing behavioral habits and lifestyle, such as through improved diet, increased exercise, and elimination of the harmful use of alcohol and tobacco, according to the WHO.

Artificial intelligence in healthcare is expected to increase worldwide from a USD 15.4 billion market-size value in 2022 to USD 208.2 billion by 2030, growing at a compound annual growth rate (CAGR) of 38.4 percent during that period, according to a report by Grand View Research. The contributing factors to this trend include the increasing embracing of precision medicine, the emerging significance of medical big data, decreasing hardware costs, and the need for reducing healthcare costs per the same report.

Using deep learning for radiology as an assistive tool is a natural fit.

Computer-aided diagnostics (CAD), an earlier form of artificial intelligence, has been used in radiology for many years for applications such as detecting breast cancer on mammography and lung nodules on chest CT scans. The earlier CAD software was coded based on domain knowledge. With the rise of AI computer vision and big data sets, the latest approach is to use deep learning to learn latent features within imaging data without hardcoding.

For this study, the researchers trained an AI deep learning algorithm, called the CXR-CVD risk model, using over 147,400 chest X-rays from over 40,600 patients in multiple centers for a controlled trial for prostate, lung, colorectal, and ovarian cancer screening sponsored by the National Cancer Institute.

The algorithm was then evaluated with chest X-rays from a second independent cohort of over 11,400 outpatients of Mass General Brigham that might be eligible for statin therapy, where around 9.6 percent had a major cardiac event over the median follow-up of 10.3 years. The researchers compared the AI algorithm’s predictive values to the established clinical standard for deciding statin eligibility.

The scientists reported in a statement that their AI deep learning algorithm was able to predict future major adverse cardiovascular events from a single chest X-ray image “with similar performance and incremental value to the established clinical standard.”