By scanning your eye, Google says its new AI can predict heart disease

20 Feb 2018

Image: Jeronimo Contreras Flores/Shutterstock

Google’s health technology subsidiary Verily has developed AI capable of spotting heart disease through an eye scan.

In the last few years alone, the advent of artificial intelligence (AI) in healthcare has transformed how we discover new means of creating new medication, but also how we are able to diagnose conditions in the human body.

So now, Google’s subsidiary Verily has revealed that its latest algorithm is able to detect whether you might be at risk of heart disease just by scanning the back of your eye.

According to The Verge, the technology is sophisticated enough that, through a scan, it can deduce a person’s age, blood pressure and whether they are a smoker.

By gathering this information, Verily said, it can determine whether the patient is at severe risk of having a heart attack, just as much as a series of tests using traditional methods could.

Publishing its findings in the journal Nature Biomedical Engineering, the researchers working with the company used the power of machine learning to analyse the datasets of 300,000 patients, including both eye scans and health details.

With this information, the AI is then able to draw patterns between someone being both older and having high blood pressure, and their eye appearing in a particular way.

Eye scans

A sample retinal image from the dataset in colour (left) and an attention map (right) indicating areas that support the prediction for high systolic blood pressure in green, which overlap with the retinal blood vessels. Image: Verily/Google

This is because the rear interior wall of the eye – known as the fundus – is covered in blood vessels, which give a strong indication of the person’s overall health, and their blood pressure in particular.

To test its accuracy, the AI was then presented with the retinal images of two patients, including someone who had a heart attack just five years prior while the other had a clean bill of health.

In 70pc of instances, the AI was able to accurately tell which of the two had experienced a heart attack, putting it slightly below the current gold standard of testing known as SCORE, which has an accuracy rate of 72pc.

While admitting that further testing is necessary before it can improve, the end goal is to develop AI capable of churning through vast amounts of records to find connections between medical data in the hope of finding conditions or solutions yet to be discovered.

Colm Gorey was a senior journalist with Silicon Republic