Can the health of your heart be revealed through a simple look into your eyes? Google is making strides in proving this theory with an algorithm that utilizes retinal images to predict risk factors associated with heart disease. By studying the anatomical features in the eyes, such as the optic disc and blood vessels, the algorithm can provide insights into age, gender, blood pressure, and smoking status — significant predictors of heart diseases.
The technology, which matches eye scans with a matrix for cardiovascular risks, has shown promising results, accurately predicting heart health in 70% of the tested cases thus far. This breakthrough AI technique developed by Google has the potential to revolutionize the way cardiovascular risks are detected, potentially eliminating the need for lengthy and conventional tests and scans.
The research behind this innovative approach was initially focused on developing a device for detecting diabetic retinopathy. Leveraging access to retinal images from large datasets in the United States and the United Kingdom, Google trained its deep learning model to generate predictions about age, gender, smoking status, and blood pressure based on the eye scans. Further validation was conducted using an additional 13,000 records from these datasets.
Building upon similar studies conducted by the University of Leeds, which explored the correlation between retinal scans and cardiac scans in over 5,000 individuals, Google’s AI system identified associations between retinal pathology and changes in heart health. By analyzing retinal patterns, the AI system could estimate the size and pumping efficiency of the left ventricle, a crucial indicator of heart disease risk. Combined with basic demographic data, such as age and sex, the AI system could predict the likelihood of a heart attack within the following 12 months.
The idea behind this innovative research emerged from Google’s collaboration with partner hospitals in India to develop an AI tool for screening diabetic retinopathy without the need for specialized personnel. Diabetic retinopathy, a complication resulting from uncontrolled blood sugar levels over time, can cause severe damage to the retina and potentially lead to blindness. Traditional diagnosis methods often rely on patients recognizing changes in their vision, which can occur at a late stage of the disease. By leveraging the power of AI, Google’s solution allowed for remote diagnosis by analyzing retinal images sent by diabetes clinics. This streamlined approach significantly reduced the burden on eye doctors, ensuring efficient and accurate screening for the condition. The AI-enabled device has already obtained the CE certification for use in Europe and has been successfully deployed by Aravind Hospital to screen 200,000 individuals in remote areas.
Dr. Rajiv Raman from Sankara Netralaya, a collaborator in the initial study, is now exploring the potential of using these retinal images to predict the occurrence of diabetic retinopathy. The challenge lies in testing the effectiveness of the algorithm with real-world data and validating its predictive capabilities. To accomplish this, data from thousands of patients, including retinal images, blood pressure readings, blood glucose levels, and cholesterol levels, must be collected. These patients will be followed for a period of two years to determine whether the algorithm can accurately predict the risk of cardiovascular diseases based on the initial retinal scans.
This groundbreaking research by Google holds immense potential in transforming the way heart disease risks are assessed and detected. By harnessing the power of AI and leveraging the insights obtained from retinal images, medical professionals may be able to identify cardiovascular risks earlier, leading to improved patient outcomes and a more proactive approach to heart health.
