Sber's artificial intelligence laboratory has created an algorithm to instantly detect the likelihood of coronavirus disease. RIA Novosti reports about it with reference to the materials of the company.
The presence of the virus is determined by symptoms and three sound patterns – voice, breathing and cough. All information about the patient in the form of sound files is collected in a spectrogram, then it is analyzed by a deep convolutional neural network. Sber's specialists clarified that the neural system was trained on the basis of more than three thousand breathing and coughing samples obtained from Russian clinics. According to experts, passing the test and getting the result takes no more than a minute.
“This is not a medical diagnostic tool, but rather a personal daily checker,” the company said. “Our model has not yet achieved the accuracy of a biological PCR test, which is not surprising, but already now has comparable characteristics,” the creators of the neural network specified. Experts promised that in the near future Sberbank will launch a special application with an algorithm that can be installed on a smartphone.
Earlier, Sberbank named the cities of Russia with the highest risk of contracting coronavirus. According to SberIndex, Voronezh and Perm are at the top of the cities with the highest risk of coronavirus infection. In terms of epidemiology, the safest are Moscow, Krasnodar and Nizhny Novgorod.