ABSTRACT: Prior work has indicated that electronic nose (enose) technology can distinguish among bacteria samples and identify patients with ventilator-associated pneumonia and rhinosinusitis. This study was performed to test the hypothesis that an enose can distinguish between biofilm-producing and non-biofilm-producing bacteria of the same species.
Biofilm-producing and non-biofilm-producing mutant strains of Pseudomonas (PA01 and Sad36) and Staphylococcus (Staph WT and Staph SrtA-B-) were incubated and then sampled by the enose. Data were evaluated by logistic regression to perform binary classification of the bacteria. Training sets and testing sets were developed in two ways: by cross-validation with a leave-1-day-out method, and then by testing the last 4 days versus the first 18 days.
By the leave-1-day-out method, logistic regression analysis of 198 samples of each bacterium determined that the training accuracy of the enose for both Pseudomonas species was 100% and the testing accuracy was 91.4% for PA01 and 88.4% for Sad36. The training accuracy of the enose for Staphylococcus species was 87.1% for Staph SrtA-B- and 88.2% for Staph WT and the testing accuracy was, respectively, 75.3 and 76.8%. By separating the first 18 days from the last 4 days, logistic regression analysis determined that the training accuracy of the enose for both Pseudomonas species was 100% and the testing accuracy was 100% for PA01 and 80.6% for Sad36. The training accuracy of the enose for Staphylococcus species was 87.7% for Staph SrtA-B- and 86.4% for Staph WT and the testing accuracy was respectively, 72.2 and 91.7%.
The enose was able to identify correctly biofilm- versus non-biofilm-producing Pseudomonas and Staphylococcus species with accuracy ranging from 72.2 to 100%, depending on the particular organism and the data analysis methodology. These results suggest that biofilms may be an in vivo marker, which the enose can use to identify patients with particularly virulent and medically recalcitrant forms of chronic rhinosinusitis.
American Journal of Rhinology 22(1):29-33. · 1.36 Impact Factor