Scientists have created a neural network that itself finds signs of a severe form of COVID-19 on CT scans. This was told in the press service by the residents of the Skolkovo Foundation – the company-developer of systems for support of medical decision-making “Third Opinion”.

Skolkovo Foundation residents taught the neural network to automatically detect hydrothorax or fluid in the lungs on computed tomography (CT) images – a symptom characteristic of severe forms of COVID-19. Artificial intelligence module “Third Opinion CT. COVID-19 ”is already used in dozens of clinics across the country and has analyzed hundreds of thousands of CT studies since the start of the pandemic.

According to the developers from Third Opinion, a mathematically accurate assessment of the volume of fluid and other signs of lung damage in COVID-19 is the most objective tool for monitoring the course of the disease. Their accurate assessment using a special program allows doctors to get an observer-independent picture of the course of the disease. Negative changes in the lungs can be “detected” at a very early stage.

The AI ​​service was developed to reduce the routine burden on radiographers. It has already been registered with Roszdravnadzor. as a medical device and entered into the Register of Russian software of the Ministry of Digital Development of Russia. This allows it to be used for non-experimental purposes in medical institutions of the country.

The development was named one of the leaders of the Experiment (in the direction of “CT COVID”) on the use of innovative technologies in the field of computer vision for the analysis of medical images in the healthcare system of the city of Moscow.