The Development and Evaluation of Gas Sensor and Electronic Nose Technology

Ventilator-associated pneumonia is one of the most lethal infections occurring in intensive care units of hospitals. In order to obtain a faster method of diagnosis, we proposed to apply the electronic nose to cultures of the relevant micro-organisms. This allows to halve the time of the analysis. We focus on the application of some chemometrical tools which enhance the performance of the method. Trilinear partial least squares (tri-PLS) regression is used to perform calibration and is shown to produce satisfactory predictions. Sample specific prediction intervals are produced for each predicted value, which allows us to eliminate erroneous predictions. The method is applied to an external validation set and it is shown that only a single observation out of 22 is being wrongly classified, so that the method is acceptable for inclusion in the clinical routine.

Serneels, S.; Moens, M.; Van Espen, P.J.; Blockhuys, F. Anal. Chim. Acta. 2004, 516, 1-5.

An algorithm is derived to compute the influence function for tri-PLS1 regression. Based on the influence function, we propose the squared influence diagnostic plot to assess the influence of individual samples on calibration and prediction. We illustrate the applicability of the squared influence diagnostic plot for tri-PLS1 to two different data sets which have previously been reported in literature. Finally we note that from the influence function, a new estimate of prediction variance can be obtained.

Serneels, S.; Geladi, P.; Moens, M.; Blockhuys, F.; Van Espen, P.J. J. Chemometrics 2005, 19, 405-411.

The aim is to evaluate the electronic nose (EN) as method for the identification of ten clinically important micro-organisms. A commercial EN system with a series of ten metal oxide sensors was used to characterize the headspace of the cultured organisms. The measurement procedure was optimized to obtain reproducibel results. Artificial neural networks (ANNs) and a k-nearest neighbour (k-NN) algorithm in combination with a feature selection technique were used as pattern recognition tools. Hundred procent correct identification can be achieved by EN technology, provided that sufficient attention is paid to data handling. Even for a set containing a number of closely related species in addition to four unrelated organisms, an EN is capable of 100% correct identification. The time between isolation and identification of the sample can be dramatically reduced to 17 h.

Moens, M.; Smet, A.; Naudts, B.; Verhoeven, J.; Ieven, M.; Jorens, P.; Geise, H.J.; Blockhuys, F. Lett. Appl. Microbiol. 2006, 42, 121-126.