The potential of an electronic nose as a screening tool for tuberculosis (TB)

<bold>Background:</bold> TB remains a global health threat, particularly in resource-limited regions where early diagnosis is challenging. There is a pressing need for an accurate, affordable, and non-invasive diagnostic tool. Breath analysis, which detects specific volatile organic compounds, holds promise for disease detection. We developed an electronic nose (e-Nose) inspired by human olfaction, augmented with artificial intelligence, to screen for TB through breath samples. <bold>Method:</bold> Conducted in diverse Indonesian cities (Yogyakarta, Klaten, Surakarta, and Papua), our study trained the e-Nose to distinguish individuals with TB from healthy subjects and those with other respiratory conditions. Afterward, we tested the e-Nose on presumptive TB patients. Participants provided breath samples into air-collecting bags, which were analyzed by the e-Nose connected to a laptop. Support Vector Machine algorithms processed the data. Diagnostic accuracy was assessed against composite reference standards, integrating clinical symptoms, bacteriological examination, chest X-ray, and follow-up. Ethical clearance was obtained, and participants provided informed consent. <bold>Result:</bold> We trained the e-Nose to 27 individuals with TB, 24 healthy controls, and 53 with other respiratory diseases (median age: 43 years; BMI: 19.5 kg/m²). We tested it on 1383 people screened for TB. No adverse events occurred. In the training phase, the e-Nose had a sensitivity of 95% (95% CI=77-100%) and specificity of 82% (95% CI=60-95%). Analysis of the testing phase is ongoing. <bold>Conclusion:</bold> In this development phase, the e-Nose had high sensitivity and specificity. Further research is imperative to validate its performance across larger and diverse populations.