THE PREDICTION OF BACTERIA TYPE AND CULTURE-GROWTH PHASE BY AN ELECTRONIC NOSE WITH A MULTILAYER PERCEPTRON NETWORK

An investigation into the use of an electronic nose to predict the cla ss and growth phase of two potentially pathogenic micro-organisms, Esc hericha coli (E. coli) and Staphylococcus aureus (S. aureus), has been performed. In order to do this we have developed an automated system to sample, with a high degree of reproducibility, the head space of ba cterial cultures grown in a standard nutrient medium. Head spaces have been examined by using an array of six different metal oxide semicond ucting gas sensors and classified by a multi-layer perceptron (MLP) wi th a back-propagation (BP) learning algorithm. The performance of 36 d ifferent pre-processing algorithms has been studied on the basis of ni ne different sensor parameters and four different normalization techni ques. The best MLP was found to classify successfully 100% of the unkn own S. aureus samples and 92% of the unknown E. coli samples, on the b asis of a set of 360 training vectors and 360 test vectors taken from the lag, log and stationary growth phases. The real growth phase of th e bacteria was determined from optical cell counts and was predicted f rom the head space samples with an accuracy of 81%. We conclude that t hese results show considerable promise in that the correct prediction of the type and growth phase of pathogenic bacteria may help both in t he more rapid treatment of bacterial infections and in the more effici ent testing of new anti-biotic drugs.