Qualitative analysis of biological tuberculosis samples by an electronic nose-based artificial neural network.

OBJECTIVE

To apply an e-nose system for monitoring headspace volatiles in biological samples from Egyptian patients with active pulmonary tuberculosis (TB) and healthy controls (HCs) and compare them with standard sputum analysis.

DESIGN

The study population comprised 260 (140 males, 120 females) newly diagnosed TB patients and 240 (120 males, 120 females) HCs matched by age and socio-economic level admitted to hospitals specialising in chest diseases in Alexandria, Behera, Giza and Damietta Governorates, Egypt. Participants provided a history of TB and underwent clinical examinations, chest X-ray, and microbiological and e-nose analyses. Biological samples (blood, breath, sputum and urine) were collected.

RESULTS AND CONCLUSION

Being a confirmed TB patient was directly proportional to e-nose 10-sensor responses. Principal component analysis clusters showed a clear distinction between TB and HC groups, with variances of 93%, 85%, 75% and 95% for blood, breath, sputum and urine samples, respectively. Overall accuracy, sensitivity and specificity of the artificial neural network (ANN) analysis for classifying samples were >99%. The e-nose successfully distinguished TB patients from HC participants for all measured biological samples with great precision. With urine samples gaining broader acceptance for clinical diagnosis, an e-nose-based ANN can be a very useful tool for low-cost mass screening and early detection of TB patients in developing countries.