Russian engineers from ID East, together with doctors from City Clinical Hospital No. 40 (infectious diseases hospital in Kommunarka), have developed Respiro acoustic screening system for respiratory diseases, which uses AI and machine learning to identify COVID-19 infection by the patient’s coughing.

The coughing is recorded on the microphone of a smartphone or a computer, after which the neural network analyzes the sound from the spectrogram and gives the result. Unlike the polymerase chain reaction (PCR) test, which gives a false negative result in 30-40% of cases, Respiro’s accuracy is more than 90% and the system is effective even in asymptomatic cases, the developers say. AI algorithms have been trained on audio recordings of coughing of more than 10,000 patients infected with COVID-19.

According to the chief physician of the medical institution Denis Protsenko, Respiro has to determine the disease and identify the need for hospitalization of the patient in the inpatient facility. Since the hospitalization itself is carried out without indications, there is a risk of nosocomial infection for patients. To date, a web service and a mobile application have been developed, they are currently in open testing. The solution will be launched in the second quarter of this year.

Original (in Russian)