Artificial intelligence from Google will immediately identify a heart attack

It would seem that a new one can be invented in the field of methods of diagnosing heart diseases, because everything has been invented long ago – know yourself to improve existing methods! However, researchers from biomedical company Verily (whose owner, by the way, is Google) were able to develop a new method for identifying cardiovascular diseases with the help of neural networks and artificial intelligence. And this method does not require taking blood and other invasive procedures, and the result is given in a few seconds.

In quality, if one may put it this way, the diagnostic material from Verily uses the image of the retina. One procedure for scanning the retina is enough to determine not only the presence of various cardiovascular pathologies, but also to reveal the person’s age, blood pressure level and to find out whether he smokes. In order to train the algorithm for diagnosis, specialists from Verily used data from 300,000 patients. These data included information on the state of the retina, as well as general information about both the patients themselves and their health status according to various indices. On the basis of the obtained data, the neural network revealed a number of patterns that can indicate the presence of cardiovascular pathology.

The main advantage of the new method is that it provides an extremely high speed of diagnostics, practically not yielding exactly to “traditional” methods. The algorithm from Verily predicts the likelihood of developing a heart attack in the next 5 years with a probability of 70%, while the most accurate SCORE method currently used (requiring, among other things, taking a blood sample) is right in about 72% of cases. Despite the prospect of development, right now it is impossible to introduce it into medical practice, as it requires further testing, debugging and monitoring of patients. Do you think that in the future AI will replace a person as a doctor? Speak out in ourtelegram chat.

Leave a Reply

Your email address will not be published. Required fields are marked *