Swaasa AI, developed by Hyderabad-based Salcit Technologies, uses machine-learning algorithms to analyse cough sounds and identify patterns associated with respiratory diseases (such as asthma, COPD, tuberculosis, fibrosis, COVID-19). In a clinical evaluation involving 355 participants, the tool reportedly achieved 97.27% sensitivity and 87.32% overall accuracy showing strong potential as a low-cost, remote, non-invasive pre-screening method.
Glimpse:
Swaasa AI, developed by Hyderabad-based Salcit Technologies, uses machine-learning algorithms to analyse cough sounds and identify patterns associated with respiratory diseases (such as asthma, COPD, tuberculosis, fibrosis, COVID-19). In a clinical evaluation involving 355 participants, the tool reportedly achieved 97.27% sensitivity and 87.32% overall accuracy showing strong potential as a low-cost, remote, non-invasive pre-screening method.
Respiratory diseases remain a leading health challenge globally but diagnosing them early often depends on equipment-heavy and expensive methods. Enter Swaasa AI: a cloud-based, acoustic analysis platform that aims to democratise lung-health screening. Developed by Salcit Technologies, Swaasa analyses cough recordings via smartphone and applies a multimodal AI model to detect anomalies linked to respiratory conditions.
In a recent multi-institutional study involving 355 participants (carried out by researchers from Andhra Medical College Visakhapatnam along with collaborators from India, the US and the UK), Swaasa flagged respiratory-disease risk with a sensitivity of 97.27% and an overall accuracy of 87.32%. The platform can further classify coughs into normal, obstructive, restrictive or mixed patterns which helps in differentiating likely conditions such as asthma, chronic obstructive pulmonary disease (COPD), pulmonary fibrosis, tuberculosis, and even COVID-19.
Previous independent validations also support Swaasa’s utility. A 2023 clinical-validation study published in a peer-review journal showed that for pulmonary tuberculosis (PTB), Swaasa achieved 90.36% sensitivity and 84.67% specificity, with overall accuracy of ~86.8% in distinguishing positive and negative TB cases. Other studies have used the platform for general lung-health screening in non-hospital settings, assessing “Lung Health Index” and detecting abnormal lung patterns (obstructive, restrictive, mixed) in populations showing promising results in community-level screening.
Because Swaasa requires no specialised hardware only a smartphone microphone and a quiet environment it holds particular promise for use in rural and remote regions, or for large-scale public-health screening. Salcit also plans integration with foundational audio-AI models (like HeAR from Google Research) to further improve detection accuracy and broaden disease coverage.
That said as with all AI diagnostics Swaasa is not meant to replace clinical diagnosis but to act as a triage and screening tool. Proper medical follow-up (spirometry, imaging, physician evaluation) remains essential for definitive diagnosis
“Analysing just a cough with AI without X-rays or spirometry and getting a reliable risk-signal can transform how we screen for lung disease in the real world.
By
HB Team
