Conventional methods of food safety testing for bacterial pathogens are accurate, but require long processing times (up to a week) to return an unequivocal determination of the presence and nature of contaminant bacteria. Biosensor-based methods, including the electronic nose (e-nose) are being researched as alternative approaches. In this paper, we present a measurement system capable of evaluating the reliability of e-nose based bacteria identification, at the genus level, based on single colonies of bacteria. Confidence measures are incorporated, which provide the user with information allowing them to better assess the reliability of any individual classification result. The system is tested with four non-pathogenic bacteria types (two from the same species, Escherichia coli). The results demonstrated classification accuracies greater than 80%. Furthermore, it is shown that higher classification accuracy (96.7%) can be achieved by repeated e-nose sampling of the same colony and using all available odor responses to characterize a sample. Single bacterial colonies are available relatively early during the conventional testing process, so further developments in this area hold the potential to shorten the testing times, thereby complementing existing methods of food pathogen testing.

Additional Metadata
Keywords Bacteria, Classification, Electronic nose, Food safety
Persistent URL
Journal Sensors and Actuators, B: Chemical
Green, G.C. (Geoffrey C.), Chan, A, & Lin, M. (Min). (2014). Robust identification of bacteria based on repeated odor measurements from individual bacteria colonies. Sensors and Actuators, B: Chemical, 190, 16–24. doi:10.1016/j.snb.2013.08.001