Representational / Getty image.
,” Google CEO Sundar Pichai said on 20 February while speaking the study in a tweet.
The analysis, published in the journal Nature Biomedical Engineering, revealed that deep learning applied to a retinal fundus image, a photograph that has the blood vessels of the eye, may forecast risk factors for heart ailments — from blood pressure to smoking status.
The algorithm which the investigators produced can even help predict the occurrence of a future major cardiovascular event on par with current measures, said Michael McConnell, Head of Cardiovascular Health Innovations at Verily in blog post.
Cardiovascular disease is the main cause of death globally and researchers know that lifestyle factors including diet and exercise in combination with genetic factors, age, ethnicity, and gender all contribute to it.
However, they do not exactly know how these factors add up in a particular individual, therefore in some patients it becomes essential to perform sophisticated tests, such as coronary calcium CT tests, to help better stratify someone’s risk for having a heart attack or a stroke, and these other cardiovascular events.
Retina picture. Google
In this study, applying heavy learning algorithms trained on information from 284,335 patients, the investigators were able to predict cardiovascular risk factors from retinal images with surprisingly high accuracy for patients from 2 separate datasets of 12,026 and 999 patients.
The algorithm could differentiate the retinal images of a smoker from this of a non-smoker 71 percent of their time, the analysis found.
“Additionally, while physicians can normally distinguish between the retinal images of patients with severe high blood pressure and normal patients, our algorithm could go further to forecast the systolic blood pressure within 11 mmHg on average for patients overall, including those with and without high blood pressure,” study co-author Lily Peng, Product Manager, Google Brain Team, stated.
“One of the exciting elements of this research is the creation of ‘focus maps’ to reveal which characteristics of the retina contributed most to the algorithm, thus providing a window into the ‘black box’ frequently related to machine learning,” McConnell, who’s also a co-author of the analysis, said.
This may give clinicians greater confidence in the algorithm, and possibly offer new insights to retinal features not previously linked to cardiovascular risk factors or potential risk, McConnell said.
The findings indicate that an easy retinal picture could one day help understand the wellbeing of an individual’s blood vessels, key to cardiovascular health.
“This is promising, but early study — more work must be done in order to develop and validate these findings on larger patient cohorts earlier this can arrive in a clinical setting,” McConnell added.