Specialists of the Novosibirsk State Technical University of NETI plan to connect machine learning methods to the analysis of recordings of acoustic signals of human lungs from a portable health monitoring device, Andrey Ivanov , head of the Department of Information Protection of NSTU NETI told Interfax.
"As the database of signal recordings accumulates, algorithms will be upgraded. When a dataset (structured set — IF) of marked data is collected, it will be possible to switch to artificial intelligence methods, train a neural network, and detect various artifacts during breathing and coughing, helping doctors track the dynamics of patients," Ivanov said.
He noted that in 2022, the specialists of NSTU NETI would develop an algorithm for calculating vital indications of patients for a portable diagnostic device created by a resident company of the technopark of Novosibirsk Akademgorodok "Medico-Biological Union".
The device, which is a small sensor that is fixed on the body with a special patch, connects to a mobile phone and measures body temperature, heart rate, respiratory movements, coughing attacks, and potentially some other acoustic parameters of breathing.
The digitized data are transferred to the database and processed on the server, if necessary; the doctor can listen to the recording in the traditional way.
"Now we are fine-tuning some functions, increasing the detection accuracy, and working on various use cases: it can be a medical institution where a doctor could track the patient ' s condition in dynamics, maybe for telemedicine, or just as a personal device so that a person tracks his condition," the agency interlocutor noted.
According to him, when developing the algorithm, the difficulty was that there are no ready-made databases of audio recordings of people's breathing, unlike, for example, X-ray databases.
"The difficulty was that it was necessary to collect data from patients with a developed device and not with traditional phonendoscopes, which requires cooperation with a medical institution and training of doctors. The recordings were then analyzed using spectral analysis methods. The difficulty also lies in the fact that each person is an individual and it is required to determine the universal settings of the device. At the moment, the option of using the device in pediatrics to monitor the breathing characteristics of children is being considered," Ivanov said.
Mikhail Losev, director of Medico-Biological Union LLC, the initiator of the project, told Interfax that remote health monitoring services are in demand in a wide variety of segments of private and public medicine.
"We plan to release a pilot series of devices at the end of the year and begin state registration of the product," he said.