A neural network has been created in Minsk to recognize pneumonia even before the onset of symptoms. This is a development of a student of the Belarusian Medical University, according to MIR 24 correspondent Olga Baranova.
The neural network analyzes up to ten X-ray images per second. Distributes them into categories: norm, pneumonia and other pathology. The accuracy of the program is above 90%. The smart system does not determine the cause of the disease. To confirm, for example, a coronavirus, you need to contact a laboratory.
“The volume was very large. We have collected about a quarter of a million images. When feeding our neural network images, the doctor receives a prediction from it - the probability of classifying this or that image. And therefore, he can prioritize and work primarily with images that show signs of pneumonia or other pathology. And accordingly, such a patient will wait less,”said Andrei Kapitonov, student of the Belarusian State Medical University.
Such a development, of course, will not replace doctors, but will speed up their work. The workload of radiographers has increased significantly. Before the pandemic, one image was analyzed for three days, now it is one. But even this waiting time can become critical for the patient.
