A new Google artificial intelligence provides cardiovascular risk by looking you in the eye
A new Google artificial intelligence provides cardiovascular risk by looking you in the eye, has learned to recognize clues on the retina, is faster and is almost as accurate as traditional examinations

Thanks to artificial intelligence, Google has developed a new system to assess the risk of developing heart disease, using a retinal scan, the innermost part of the eye. The results obtained so far are very promising and comparable to those obtained with traditional diagnostic systems. The research was conducted by Google in collaboration with Verily, a subsidiary company that deals with new technologies to be applied in healthcare. The study was published in the scientific journal Biomedical Engineering and its first preliminary version was published last fall.

Analyzing an image of the patients' retina, the algorithm can calculate the patient's age and blood pressure, in addition to detecting any habits - such as smoking - that involve a significant increase in risk, especially in predisposed individuals. By combining these data derived from the ocular scan, the system is able to make a prediction of the risk that proved to be as reliable in the tests as that achieved with traditional methods.

The algorithm was trained with the classic methods of "machine learning", thus offering an enormous amount of data to recognize trends. The system analyzed 300,000 medical reports from as many patients, for whom other medical information was collected in addition to scans of their retinas. In this way he learned autonomously to associate particular characteristics in the retina with other indicators, which contribute to calculating cardiovascular risk. The retina is crossed by an intricate series of vessels, from whose analysis many things can be understood about a patient's health.
A new Google artificial intelligence provides cardiovascular risk by looking you in the eye

After machine learning, researchers from Google and Verily tested their algorithm by presenting two images of as many retinas: one from a patient who had suffered from a cardiovascular problem in the last 5 years, the other from a healthy patient. In 70 percent of the cases, the algorithm was able to correctly indicate which of the two images belonged to the subject with the highest cardiovascular risk. The result is remarkable, considering that the classical methods of evaluation work in 72 percent of the cases (on equal terms with the experiment made with artificial intelligence). Traditional tests are also more invasive and require more time, because they require, for example, a blood sample.

Doctors have long known that with the analysis of the retina we can understand many things about the health of a patient, from a cardiovascular point of view. The algorithm can then be used to make diagnoses much faster and extend them to more patients, making it possible to more accurately control the population at risk. The system is of course in its infancy and, although promising, will have to go through other tests before becoming part of the common practices to make the diagnoses.

Post a Comment

Powered by Blogger.