Alphabet’s health division, Verily, has a number of ongoing projects that leverage technology and data science to improve healthcare. The latest applies machine learning to retinal images to identify the risk factors of cardiovascular disease.
In a paper published today (via The Verge), Verily and the Google AI teams detail their work analyzing blood vessels at the back of the eye to predict risk factors, like blood pressure and smoking, associated with cardiovascular disease.
These retinal fundus images — once analyzed by deep learning models — can determine age, gender, smoking status, and systolic blood pressure. In turn, that data is correlated to help determine how those factors influence “major adverse cardiac events,” like heart attacks.
Verily trained these models using data from nearly 300,000 patients, with the system then associating these factors together. The algorithm was ultimately able to determine which patients eventually developed an ailment, with 70% accuracy. In comparison, the current standard based on blood tests is able to determine with a 72% rate.
In addition to predicting the various risk factors (age, gender, smoking, blood pressure, etc) from retinal images, our algorithm was fairly accurate at predicting the risk of a CV event directly. Our algorithm used the entire image to quantify the association between the image and the risk of heart attack or stroke.
This system can generate attention maps that visually highlight how the algorithm is arriving at its conclusion. In the example above, the attention map indicates areas that correlate to various factors in green.
One of the exciting aspects of this study is the generation of “attention maps” to show which aspects of the retina contributed most to the algorithm, thus providing a window into the “black box” often associated with machine learning. This can give clinicians greater confidence in the algorithm, and potentially provide new insights into retinal features not previously associated with cardiovascular risk factors or future risk.
The act of taking a picture of one’s retina is much less invasive and cheaper than current tests like coronary calcium CT scans. It opens up a wealth of possibility in regards to preventive care and screening for heart disease.
The opportunity to one day readily understand the health of a patient’s blood vessels, key to cardiovascular health, with a simple retinal image could lower the barrier to engage in critical conversations on preventive measures to protect against a cardiovascular event.
A promising start, this research is still in its early days, with Verily noting that “more work must be done to develop and validate these findings on larger patient cohorts before this can arrive in a clinical setting.”