A cross sectional feasibility study to evaluate the usability and efficacy of Swaasa AI platform for rapid respiratory health assessment

Journal article

Analysing cough sounds is vital in pulmonary medicine. Recently, AI tools are being trained to analyse the acoustic signals of cough sounds so that more cases can be quickly tested, thereby reducing the patient load on primary healthcare systems. In this paper, we evaluate ”Swaasa”, our AI-based platform for rapid respiratory screening, highlighting its efficacy and ease of use. We applied our trained classifier to catch underlying pathologies from cough sound data collected from diverse sources. We then used a pattern classifier to identify specific respiratory conditions based on cough sound patterns. We tested the robustness of our methods by comparing our results with that of a pulmonary physician in 355 cases and show that Swaasa correctly predicted associated risk in 87.32% of those cases. Our platform has a sensitivity of 97.27% with a Positive Predictive Value (PPV) of 88.54%, giving us the potential to revolutionise disease screening, especially for large populations and in isolated rural areas. Our rapid and easy-to-use Software as a Service (SaaS) solution efficiently diagnoses and conserves resources, improves patient outcomes, and establishes a comprehensive and accessible healthcare framework.

Authors: Lotheti, SK; Vissamsetti, H; Pamarthi, K; Bhimarasetty, DM; Rudraraju, G

• Publication status: Published
• Peer review status: Peer reviewed
• Publisher: Nature Research
• Journal: Scientific Reports
• Volume: 15
• Issue: 1
• Article number: 38379
• Publication date: 2025-11-03
• Acceptance date: 2025-09-26
• DOI: 10.1038/s41598-025-22215-8
• EISSN: 2045-2322
• ISSN: 2045-2322
• Language: English
• Keywords: Respiratory health; Convolutional neural network (CNN); Cough signature; Machine learning; Feedforward artificial neural network (FFANN)