Peter the Great St. Experts from St. Petersburg Polytechnic University (SPbPU) have developed the most accurate neural network to detect COVID-19 using X-ray and CT images of the lungs. The accuracy of the neural network was found to be 99.23%, this issue was conducted in St. Petersburg Polytechnic University, socialbites.ca was told.
Today, PCR is the standard for diagnosing the disease, but the analysis procedure has many disadvantages. First of all, it takes 4-6 hours to get the results. Secondly, there is a risk of false negative results, especially in the early stages of the disease. Therefore, searches for early diagnosis of the disease continue.
Imaging of the lungs using X-ray or CT is a good addition to PCR testing. These procedures also help rule out coronavirus-related pneumonia. However, interpreting the results of an X-ray or CT image requires the utmost attention of highly qualified specialists.
To facilitate the image interpretation process, St. Researchers at St. Petersburg Polytechnic University created a neural network that identifies symptoms of Covid-related pneumonia based on computed tomography images.
“Our model allows us to quickly and accurately detect the presence of COVID-19 or other types of pneumonia on CT images of the lungs. This can help doctors make faster and more accurate diagnoses, reducing the burden on medical staff. Our program is important, first of all, in terms of improved interpretation accuracy different from analogues,” one of the developers of the neural network, engineer of the world-wide Scientific Center “Advanced” Scientific and Technological Complex “Mathematical Modeling and Intelligent Control Systems”. Digital Technologies” SPbPU Dib told socialbites.ca.Ali.
To train the neural network, the scientists used a series of COVID-19 computed tomography slices obtained from different countries. It includes more than 7,500 images of lungs affected by coronavirus-related pneumonia, more than 2,500 images of lungs with community-acquired pneumonia, and approximately 7,000 images of healthy lungs.
The results of fourfold cross-validation of the new neural network proved its accuracy and efficiency. According to the researchers, the model has great potential to accurately and quickly diagnose COVID-19 using CT images.
Previous scientists was created Neural network to detect deepfakes in photos and videos.