Electronic lung auscultations in pulmonology: from clinical skill to digital diagnostics

Lung auscultation remains a fundamental method for the clinical diagnosis of respiratory diseases; however, its effectiveness is limited by subjective interpretation and interobserver variability (kappa coefficient (Κ) 0.4—0.6). Advances in digital technologies have led to the emergence of electronic auscultation as a tool for the objectification and standardization of auscultatory data. Objective. To systematize data on modern electronic stethoscopes, open-access respiratory sound databases, artificial intelligence (AI) algorithms, and telemedicine capabilities, with an assessment of their clinical applicability. Material and methods. A literature review was conducted using publications retrieved from PubMed/MEDLINE, Embase, IEEE Xplore (Institute of Electrical and Electronics Engineers), and eLIBRARY.RU databases for the period 2000—2025. Results. Modern devices perceive an extended frequency range (up to 2000 Hz), provide recording, visualization, and transmission of data to medical information systems. Deep learning algorithms — convolutional neural networks, long short-term memory networks, and vision transformers — demonstrate classification accuracy of respiratory sounds ranging from 79% to 99% on public datasets. Key limitations identified include the lack of unified data storage standards, inter-device variability in amplitude—frequency characteristics, insufficient clinical validation, and limited integration with international standards for medical data exchange and the Unified State Health Information System. Conclusion. Electronic auscultation is a key component in transition to digital respiratory medicine. Its full-scale implementation requires the development of regulatory standards, creation of multicenter national respiratory sound databases, and clinical validation of AI algorithms.