USE OF NEURAL NETWORKS FOR QUANTITATIVE MEASUREMENTS IN ION MOBILITY SPECTROMETRY (IMS)

The use of a neural network (NN) to derive quantitative information from the mobility spectra of dimethylformamide, bromine, and hydrogen fluoride is described. Unlike traditional methods, where the areas or heights of known and identified peaks are used for calibration, employment of NN does not require detailed knowledge of the ion chemistry of the measured system. Efficient NN training algorithms were employed, allowing quick learning with hundreds of input channels. It was also found that the size of the required training set was much smaller than expected from theoretical considerations, due to the interactions between input channels. Thus, this technique may be easily implemented in online or offline analysis of IMS spectra.