Software devised by Google is able to predict if your blood pressure is too high or even if you could have a heart attack in the near future, among other conditions, just by examining your eyes. What’s more, its managers say in a recent statement that this technology, which uses artificial intelligence algorithms, could favor the development of new techniques for medical diagnosis based on the analysis of images of the retina.
In a study published in the journal Nature Biomedical Engineering , a team of computer scientists from the research division of Google and Verily, one of its subsidiary companies, dedicated to the study of life sciences, explains that this computer tool was capable of of noting several cardiovascular risk factors that, to date, it was not believed that it was possible to quantify in this type of snapshots, such as the subject’s age –with an error of just three years–, their sex, whether or not they smoked. which successfully determined 71% of the time – his blood pressure and the possibility that he would suffer a serious heart problem in the next five years, with an efficiency of 70%.
Almost as good as a doctor
The authors of the trial point out that this percentage is only slightly lower than that achieved with some of the methods currently used to predict cardiovascular risk, such as SCORE, which, in addition to an examination of the patient’s habits , requires a blood test. However, for a program to carry out this association is an especially complex task due to the great variety of patterns, colors and shapes that it must take into account. To do this, they used a machine learning model that,to train, he examined data from 284,335 patientsand he studied certain anatomical features, such as the optic disc – a structure located in the center of the retina – or the blood vessels.
In the future, this technique could be used to develop applications that allow doctors to detect in advance and non-invasively the appearance of possible heart ailments. In Spain, for example,they cause three out of ten deaths, which makes them the number one cause of death. However, experts caution that before this novel technology can be introduced, it must be thoroughly tested by other teams.
Health is one of the areas that artificial intelligence could revolutionize in the coming years. Thus, some identification systems that already use it are capable of detecting the presence of tumor masses in medical images more efficiently than humans. In 2017, Google announced that its artificial intelligence could diagnose cancer faster than a pathologist and with greater reliability: its algorithm is correct 89% of the time compared to 73% of specialists.
Referencia: Prediction of cardiovascular risk factors from retinal fundus photographs via deep learning. Ryan Poplin et al. Nature Biomedical Engineering(2018). DOI: 10.1038 / s41551-018-0195-0
Image: On the left, an image of the human eye fundus. On the right, how Google’s algorithm captures it. In green, the regions that are most useful for predicting blood pressure (USAF / Google).
