Detection of Mycobacterium tuberculosis (TB) in vitro and in situ using an electronic nose in combination with a neural network system.

The use of volatile production patterns produced by Mycobacterium tuberculosis and associated bacterial infections from sputum samples were examined in vitro and in situ using an electronic nose based on a 14 sensor conducting polymer array. In vitro, it was possible to successfully discriminate between M. tuberculosis (TB) and control media, and between M. tuberculosis and M. avium, M. scrofulaceum and Pseudomonas aeruginosa cultures in the stationary phase after 5-6h incubation at 37 degrees C based on 35 samples. Using neural network (NN) analysis and cross-validation it was possible to successfully identify 100% of the TB cultures from others. A second in vitro study with 61 samples all four groups were successfully discriminated with 14 of 15 unknowns within each of the four groups successfully identified using cross-validation and discriminant function analysis. Subsequently, lipase enzymes were added to 46 sputum samples directly obtained from patients and the head space analysed. Parallel measurements of bacterial contamination were also carried out for confirmation using agar media. NN analysis was carried out using some of the samples as a training set. Based on the NN and genetic algorithms of up to 10 generations it was possible to successfully cross-validate 9 of 10 unknown samples. PCA was able to discriminate between TB infection alone, the controls, M. avium, P. aeruginosa and a mixed infection. These findings will have significant implications for the development of rapid qualitative systems for screening of patient samples and clinical diagnosis of tuberculosis.