AI Trained to Detect Four Lung Diseases Through Exhaled Breath

Specialists from the Institute of Personalized Cardiology at Sechenov University have trained an artificial intelligence system to detect four serious lung diseases by analyzing the chemical composition of exhaled breath. The conditions include chronic obstructive pulmonary disease (COPD), bronchial asthma, and two rare and severe disorders: cystic fibrosis and lymphangioleiomyomatosis.

Some of these diseases have traditionally required complex molecular and genetic testing. The new Russian technology could offer a faster and non-invasive alternative. The findings were published in the International Journal of Molecular Sciences.

How does the new method work?

When people breathe, they exhale hundreds of volatile organic compounds. As pathological processes develop in the body, the concentrations of these substances can change, creating a distinctive chemical “fingerprint.” The researchers developed a specific exhaled-breath profile for each of the four diseases.

For the patient, the procedure is remarkably simple: they only need to breathe into a special medical mouthpiece for a short period. The sample is analyzed directly using high-precision equipment known as a proton-transfer-reaction mass spectrometer. The instrument breaks the exhaled breath down into its individual components and measures the levels of volatile molecules, while AI performs the main diagnostic analysis, according to Nauka Rossii.

The neural-network algorithms were trained on large volumes of clinical data and learned to correlate the concentrations of dozens of chemical compounds. A diagnostic profile may include 15 to 30 or more biomarkers, which the system combines into a single diagnostic pattern.

If the chemical profile of a person’s breath matches a pattern associated with a particular disease, the AI system flags a high probability that the condition is present.

The researchers believe the approach could eventually make the diagnosis of certain lung diseases faster and less invasive. However, AI-based breath analysis is currently a diagnostic research technology and would require further validation before it could replace established clinical tests.