Health IT

Google trains machine learning algorithm to use eyes to learn about the heart

If the eyes are the window to the soul, Google thinks it's proven that images of the retina can offer a window on the state of the heart.

Human eye anatomy, retina, optic disc artery, and vein etc. Getty Images.

If the eyes are the window to the soul, Google thinks it’s proven that the retina can offer a window to the state of the heart. A study which used data from 284,335 patients to train Google’s deep learning algorithm, a variation of machine learning, to assess and predict cardiovascular disease risk based on changes in retinal images revealed a relatively high rate of accuracy.

Currently, when doctors assess risk for cardiovascular disease, they take into account information such as age, sex, smoking, blood pressure and cholesterol levels from a blood test.

The study was published in Nature Biomedical Engineering, and What’s interesting is that the blog entry claimed that the algorithm could detect whether the patient smoked or not and assess their systolic blood pressure based on image analysis. Just by analyzing the images, the algorithm could distinguish the retinal images of a smoker from a non-smoker 71 percent of the time, according to the study. In another example, when the algorithm processed images of patients who would suffer a heart attack within five years and those who did not, the algorithm could make the correct assessment 70 percent of the time.

Despite the proof that machine learning can be accurate for interpreting risk factors based on medical images, one of the greatest challenges for wider adoption of machine learning in medicine is showing how these algorithms arrive at their conclusions. The study also addressed this question, as the blog entry points out:

…we opened the “black box” by using attention techniques to look at how the algorithm was making its prediction. These techniques allow us to generate a heatmap that shows which pixels were the most important for a predicting a specific CV risk factor. For example, the algorithm paid more attention to blood vessels for making predictions about blood pressure, as shown in the image above. Explaining how the algorithm is making its prediction gives doctor more confidence in the algorithm itself. In addition, this technique could help generate hypotheses for future scientific investigations into CV risk and the retina.

What’s compelling about the study is that it could point the way forward to developing more accurate ways of using anatomical changes that can be picked up in medical images to predict disease risk. In future studies, the researchers said they plan to explore the effects of interventions such as lifestyle changes or medications on risk predictions.

Photo: lightmemorystock, Getty Images 

Shares1
Shares1