Making a smart instrument: Chemometric methods applied to ion mobility spectrometry for pattern recognition and feature extraction.

摘要

Chemometric methods are applied to Ion Mobility Spectrometry (IMS) with the goal of making the next generation of IMS instruments capable of intelligent data interpretation. The first two projects were modifications to SIMPLISMA (SIMPLe to use Interactive Self-Modeling Mixture Analysis), a multivariate feature extraction method. The first modification was the implementation of SIMPLISMA in near-real time. The second modification to SIMPLISMA was to improve feature extraction capabilities on IMS data. The, third and final project was the comparison of IMS compression methods for direct use with SIMPLISMA on the compressed data. The near-real time implementation of SIMPLISMA called RSIMPLISMA is able to be implemented four times faster then previous versions of SIMPLISMA. RSIMPLISMA is able to analyze data faster then current instruments generate data. The enhanced feature extraction obtained from SIMPLISMA in PSIMPLISMA is able to extract features from IMS data sets when two components in the data have very similar profiles. It is possible to extract features from wavelet compressed data directly with SIMPLISMA. The wavelet transform retains enough time domain information in the transformed domain to allow SIMPLISMA to distinguish IMS features. Fourier compressed data does not have enough resolution of IMS features in the frequency domain to allow SIMPLISMA to extract meaningful IMS features. The application of SIMPLISMA on compressed data set allows a fast and cleaner extraction of variables for continued analysis and comparison.