The promise of Big Data is that if you can aggregate enough information, artificial intelligence (AI) algorithms can sift through it and find correlations that a human mind might never find in a lifetime of searching. We’ve covered how Google’s AI labs have learned to diagnose cancer from medical images. Well, Google is at it again, this time detecting signs of cardiovascular disease.

They say the the eyes are the windows on the soul. According to Google scientists, they may also be the window that lets us look into the future and predict heart disease. The researchers used retina photographs of more than a quarter million patients to train a deep learning system. They were then assess cardiovascular risk factors with impressive accuracy. When given images from two patients, the system was able to predict which one would have a heart attack within the next five years with 70% accuracy. The system could also estimate systolic blood pressure within 11 mmHg on average, just from a static retina image. It was 71% accurate at identifying whether or not the patient was smoker.

The researchers were also able to find out what the deep learning system was looking at as it made its judgments. As might be expected, blood vessels were the key indicators for blood pressure estimates. This new ability to assess a patient’s heart health simply from a retina image is certainly important, but that’s the smaller part of this story. The big deal is that this represents a new, effective way to use massive amounts of historical health data to come up with new ways to identify disease, often long before any symptoms become apparent.