AI sniffs out tuberculosis in breath’s chemical signature

Physicists from Botswana and South Africa have used machine learning to detect tuberculosis by analyzing the chemicals in a person's breath, Physics World reports. The team studied volatile organic compounds from patients with active TB, drug-resistant TB, and healthy volunteers, feeding nearly two thousand spectral features into four machine-learning algorithms. A linear support vector machine distinguished between the three groups with ninety-three percent accuracy. The breakthrough also revealed something practical: the most diagnostic molecules are clustered within a specific twenty-minute window of the chemical analysis, meaning future diagnostic tools could be simplified and made faster. Non-invasive breath testing could be especially transformative in resource-limited settings, where traditional TB diagnostics—slow sputum cultures and laboratory backlogs—delay treatment for a disease that remains one of the world's deadliest infections.

Listen to this story

Hear this and more stories in a personalized audio briefing.

Open The Chonkerton