Published: Tue, February 20, 2018
Medicine | By Daryl Nelson

Google Uses Deep Learning To Help Predict Heart Disease Risk

Google Uses Deep Learning To Help Predict Heart Disease Risk

The paper shows that deep learning can extract new knowledge from retinal fundus images and is able to identify cardiovascular risk factors.

To "train" the algorithm, Verily's scientists used machine learning to study medical data from nearly 300k patients - some with cardiovascular issues, some without.

The procedure for this involves analyzing blood vessels in a region of the eye called the retinal fundus, Newsweek reports. According to USA Today, by looking at the images, 70% of the time Google's AI was able to accurately predict which patient would experience a heart attack or other major cardiovascular event within five years. Learning whether you are prone to heart disease can prevent millions of deaths every year as heart disease is the top killer worldwide.

This new algorithm is fairly accurate at predicting the risk of a cardiovascular event directly by scanning the eyes. Their device looks at scans of the retina or the back layers and membranes of the eye and can then predict features such as the person's age, his or her blood pressure and also if they smoked. "In addition, while doctors can typically distinguish between the retinal images of patients with severe high blood pressure and normal patients, our algorithm could go further to predict the systolic blood pressure within 11 mmHg on average for patients overall, including those with and without high blood pressure", notes Google in its blog post. The software can also predict the risk of stroke and if a person is affected by high blood pressure. Michael V McConnell, head of Cardiovascular Health Innovations at Verily, said that the research needs more work and a larger patient database to validate these findings before it's ready for clinical testing.

Kate Middleton, her Baby Bump & Prince William attend 2018 BAFTAs
She said: "It's more likely we'll see an alien onscreen than we'll see an Asian woman at the moment, which is disgraceful". Others defended her neutral stance - including polarizing British journalist and TV personality Piers Morgan.

While most of these factors can be ascertained by simply asking the patient, other factors such as cholestrol require drawing blood. All of these factors are important predictors of cardiovascular health.

"The caveat to this is that it's early, (and) we trained this on a small data set", says Google's Lily Peng, a doctor and lead researcher on the project.

As Peng notes, opening the black box to explain how predictions are made should give doctors more confidence in the algorithm.

Google used models based on data from 284,335 patients. In future studies, the researchers said they plan to explore the effects of interventions such as lifestyle changes or medications on risk predictions.

Like this: