Almost one hundred percent
Artificial intelligence is able to detect COVID-19 by two blood indicators
Russian scientists have developed an algorithm based on artificial intelligence, which allows you to determine in just a few minutes whether a person is sick with coronavirus. The program analyzes eleven blood parameters, according to which it detects an infection. At the same time, a combination of just two of them is enough to make a diagnosis with 99.8% accuracy. The development will help doctors to quickly and inexpensively test for COVID-19 in hospitals and beyond, since the algorithm can be used in compact wearable devices. The results of the study, supported by a grant from the Russian Science Foundation (RNF), are published in the journal Sensors (Velichko et al., Machine Learning Sensors for Diagnosis of COVID-19 Disease Using Routine Blood Values for Internet of Things Application).
Scientists from Petrozavodsk State University (Petrozavodsk) with foreign colleagues have developed an algorithm based on artificial intelligence, which allows you to quickly determine whether SARS-CoV-2 infection is present in the human body. The algorithm analyzes eleven biomarkers − signs by which the condition of the body can be assessed are in the blood and, comparing these indicators with each other, determines with an accuracy of more than 99% whether the patient is sick.
To train the artificial intelligence system, the authors used open databases on biomarkers in people tested for COVID-19. These data sets were collected by the Turkish Research Hospital in 2021. Using the database, scientists compared 51 biomarkers in patients with positive and negative test results for COVID-19. Artificial intelligence revealed 11 signs that clearly differed in sick and healthy people. Among them were the level of cholesterol and its transferring molecules, as well as the amount of hemoglobin in red blood cells − red blood cells.
Next, the authors checked how many signs would be enough to unambiguously determine the patient's diagnosis. It turned out that if an artificial intelligence system evaluates data on all 11 biomarkers, the accuracy of its predictions is 100%. If only for any two or three of them − 99.8% and 99.9%, respectively. This suggests that even on the basis of a small amount of data, the system detects an infection with high accuracy.
"Our proposed algorithm will allow us to quickly test patients with suspected COVID-19 in hospitals. We suggest using it not only in stationary computers, but also in human-wearable sensors (healthy lifestyle gadgets), with the help of which it will be possible to conduct rapid testing outside medical institutions. In the near future, it will not be a medical device, but a device for household use and prevention of the spread of infection," says Dmitry Korzun, PhD in Physics and Mathematics, Associate Professor of the Department of Computer Science and Mathematical Support of Petrozavodsk State University, the head of the project supported by a grant from the Russian Academy of Sciences.
Portal "Eternal youth" http://vechnayamolodost.ru