AI for coronavirus diagnosis
Artificial intelligence trained to detect coronavirus pneumonia on X-rays
Over the past few weeks, artificial intelligence has detected more than 400 cases of pneumonia caused by a new type of coronavirus.
Previously, the technologies of the Third Opinion project (a resident of the Skolkovo Foundation and a finalist of the prestigious business accelerator) were tested in Moscow and the regions to detect signs of blood cancer and tuberculosis, and now they are considering using them to diagnose coronavirus pneumonia.
In order not to miss the disease in polyclinics, neural networks first of all analyze digitized X-rays of the lungs. The neural network gets access to the images and draws up a conclusion, where it indicates the presence or absence of four signs characteristic of coronavirus pneumonia, marks these signs on the "heat map" and calculates the probability of a COVID-19 diagnosis. The accuracy of diagnosis for three of the four signs reaches 94%.
In the presence of minimal suspicions, neural networks switch to the analysis of other signs, monitors blood parameters, for example, records excess leukocytes and other signs. Thus, taken together, these measures make it possible not to miss COVID-19 at an early stage.
As a rule, patients with coronavirus enter the hospital already with extensive lung lesions. At the same time, polyclinics cannot quickly detect the disease, as they are not designed for simultaneous urgent X-ray diagnostics of thousands of people. In such conditions, neural networks reduce the time for making a single diagnosis.
In addition, neural networks are able to track any changes in the condition of patients and signal them to doctors. To do this, artificial intelligence similarly examines the results of computed tomography. At the same time, video monitoring algorithms detect alarming signs in the movements of patients, observing them with the help of cameras installed in the wards.
The data obtained allow doctors to understand in advance, to understand that the patient needs to be connected to a ventilator or the cessation of pulmonary ventilation, and possibly a change in drug treatment tactics. The probability of death of the patient is significantly reduced.
The developers of the Third Opinion project expressed their readiness to expand assistance and connect their neural networks to work in metropolitan clinics and other regions. The technology has no technical restrictions on the number of processed images.
<…>
All relevant information on the coronavirus situation is available on the stopcoronavirus websites.rf and access to all.rf. Coronavirus hotline: 8 (800) 2000-112. In addition, the information is available via the hashtag #MyPlace.
Portal "Eternal youth" http://vechnayamolodost.ru