A fast mass spectrum screening technique for volatile organic compounds based on parallel artificial neural networks

A technique for screening mass spectra for the presence of volatile organic compounds (VOCs) is developed using probabilistic neural networks. A parallel neural network filter is designed to recognize benzene, toluene, ethyl benzene, and o-xylene in gas chromatography-mass spectrometry (GC-MS) chromatograms of VOC mixtures. The filter trained rapidly and was evaluated by analyzing a variety of VOC combinations. The performance of the network offers some significant advantages over the traditional GC-MS data processing techniques such as ion extraction and compound library searching. Advantages include speed, selectivity, and the ability to discriminate between overlapping compounds.