AI model predicts a patient’s risk of death—and all it needs is an X-ray

Researchers have developed an advanced AI model capable of using chest X-rays to predict a person’s long-term all-cause and cardiovascular mortality risk, sharing their findings in JACC: Cardiovascular Imaging.

The model, a convolutional neural network (CNN), was designed to “read” chest X-rays and then calculate the biological age (CXR-age) of the individual in question. Biological age, the authors explained, can be a more effective resource for clinicians than chronological age.

“An advantage of framing risk in terms of biological age is that it can be easier for patients to grasp than risk score probabilities,” wrote Vineet K. Raghu, PhD, a radiologist at Massachusetts General Hospital and Harvard Medical School, and colleagues. “In this study, biological age was defined as age-normalized mortality risk. For example, a chronologically 70-year-old individual with a biological age of 65 years has similar risk of mortality and expected longevity as the average 65-year-old.”

The team developed its CNN using chest X-rays from more than 116,000 patients, validating the model with data from two independent sources. Overall, the team found that a five-year increase in biological age, as determined by the CNN, was associated with a higher risk of all-cause mortality than a five-year increase in the person’s chronological age. Similar findings were seen in increases in cardiovascular mortality, suggesting that using this advanced AI model to measure a person’s biological age is more helpful for determining the risk of all-cause or cardiovascular mortality than simply going by their chronological age.

“Chest x-rays are among the most common tests in medicine,” the authors wrote. “A future implementation could compute CXR-Age from routine chest X-ray images, to give a better estimate of an individual’s biological age.”

Could these findings lead to improve age-based risk scores? Raghu and colleagues hope to find out in the years ahead.

Click here for the full analysis. Also, the team will be sharing its AI model as free, open-source software.