Related Experiment Videos
Classification of ion mobility spectra by functional groups using neural networks.
1Department of Chemistry and Biochemistry, Eastern Washington University, Cheney 99004, USA. suzanne.bell@mail.ewu.edu
Analytica Chimica Acta
|September 7, 2001
Summary
Neural networks can accurately classify chemicals using ion mobility spectra. This automated system shows promise for rapid chemical identification, even with basic equipment.
Area of Science:
- Analytical Chemistry
- Computational Chemistry
Background:
- Ion mobility spectrometry (IMS) provides valuable chemical information.
- Automated chemical identification is crucial for various applications.
Purpose of the Study:
- To develop and evaluate a neural network model for classifying chemicals based on whole ion mobility spectra.
- To assess the performance of the model across different concentrations and chemical classes.
Main Methods:
- Training neural networks on a database of 3137 ion mobility spectra from 204 chemicals.
- Implementing spectral pre-processing and optimizing network parameters.
- Utilizing a two-tier classification approach (class identification followed by individual chemical elimination).
Main Results:
- Achieved a 0.91 fraction of successful classifications by functional group across various concentrations.
- The neural network effectively utilized features like drift times, peak characteristics, and intensities.
- A two-tier system resulted in only one false positive out of 161 test spectra.
Conclusions:
- Ion mobility spectra contain sufficient detail for automated chemical identification systems, even with low-resolution instruments.
- Neural networks offer a robust and effective method for analyzing IMS data.
- The developed system demonstrates high accuracy and potential for practical applications.