next up previous contents
Next: Compound analysis by ombining Up: Computational Information Processing Previous: Tree Level Measurements in

Mass Spectrometry Based Cola Beverage Classification

Researchers: Jukka Heikkonen


In this work the classification of different cola drinks was carried out on the basis of drinks' mass spectra measured by membrane inlet mass spectrometry. The study was carried out with VTT, Technical Research Centre of Finland.

Membrane Inlet Mass Spectrometry (MIMS) is an efficient tool for determination of organic compounds from samples of interest. In the MIMS method a thin membrane, usually a silicone membrane, is used as an interface between a sample and the vacuum of a mass spectrometer. Organic compounds, usually volatile and non-polar or medium polar, absorb into a membrane, diffuse through it and desorb into the vacuum of the mass spectrometer. Since sample molecules enter the mass spectrometer at the same time, there is no chromatographic separation of different compounds, and a composite mass spectrum is obtained.

In this study MIMS with different computational methods is applied to the classification of 13 cola beverages. For each cola sample the measured MIMS spectrum represents a 150 dimensional spectrum for analysis. For quantitative diversity analysis of the cola classes the mean pairwise difference (MPD) method was applied to obtain information about the suitability of the resulting cola spectra for classification. MPD showed that the MIMS is able to provide a compact feature space for the cola beverage classification. Finally, the sample cola beverages were classified using the k-Nearest Neighbor (kNN) classifier. The analytic results showed that the sample cola beverages can be effectively separated from one another via MIMS spectra.


 
Figure 4
Figure 4: Cola beverage samples displayed according to their mass spectra's first two prinpical components.


next up previous contents
Next: Compound analysis by ombining Up: Computational Information Processing Previous: Tree Level Measurements in
www@lce.hut.fi