Image: Pixabay

Google and its technology subsidiary focused on health, Verily, have come up with a new way to assess a person’s risk of heart disease: through machine learning. They have developed a software capable of analyzing the scans of the back of a patient’s eye.
Through these records, the algorithm is able to accurately deduce data, including a patient’s age, blood pressure or even whether they smoke or not. In fact, and as they say, the software makes it faster and easier for doctors to analyze a patient’s cardiovascular risk, since it does not require a blood test.
The article on this amazing development has been published in Nature this week. Behind it is Luke Oakden-Rayner, a medical researcher at the University of Adelaide who specializes in the analysis of machine learning. According to Rayner: “The work was solid and shows how AI can help improve existing diagnostic tools. They are taking data that has been captured for a clinical reason and they are getting more out of what we are doing now. Instead of replacing doctors, AI is trying to expand what we really can do.”
To train the algorithm they used the machine learning tools with which to analyze a set of medical data of almost 300,000 patients. This information included eye scans and general medical data. As with all deep learning analyzes, neural networks were used to extract this information from the patterns, learning to associate telltale signs in eye exams with the measures needed to predict cardiovascular risk.

The fact of analyzing through the eyes to find out the health of a person is not trivial, it is based on a set of already established research. The inner back wall of the eye is filled with blood vessels that reflect the general health of the body. By studying their appearance with the camera and the microscope, doctors can infer things like blood pressure, age, and whether they smoke or not, which is an important step in predicting cardiovascular health.

For Google, the work points the way towards a new paradigm driven by artificial intelligence towards scientific discoveries. It is true that the idea of ​​a doctor in the form of AI generating new diagnoses without human supervision is a somewhat distant perspective, it will probably be decades before we see it in the future. However, the research suggests that the idea is far from far-fetched.