Quantitative Multicomponent Spectral Analysis Using Neural Networks
Recent developments in dye chemistry and modern optics have made simultaneous measurement of multiple intracellular ion species possible1. Analysis of optical spectra from intact tissues is complicated because of the presence of multiple components. Quantitative descriptions of these components are required to apply these techniques to analytical chemistry and cellular physiology. Ratio methods2 have been applied to many areas of quantitative optical signal measurement because the quantitative data is independent of dye concentration, path length and light source intensity. However, it is known that the ratio method can give erroneous results3 without special attention to the choice of the optimal wavelengths for these methods. Full spectra carry all the information needed for qualitative and quantitative analysis. Methods like principal component regression (PCR) and partial least-squares (PLS), which have been used for full spectra calibration in most chemometrics literature4,5, or multicomponent stripping for image applications 6,7,8 need heavy computation times to perform their optimization processes.
KeywordsArtificial Neural Network Artificial Neural Network Model Weight Matrix Output Node Principal Component Regression
Unable to display preview. Download preview PDF.
- 11.C.W. Lin, J.C. LaManna, and Y. Takefuji, Quantitative measurement of two-component pH-sensitive colorimetric spectra using multilayer neural networks, Biol. Cvbern. (in press)Google Scholar
- 15.D.E. Rumelhart, and J.L McClelland, “The PDP Research Group: Parallel Distributed Processing”, MIT Press, Cambridge (1988)Google Scholar
- 22.S. Bassnett, L. Reinisch, and D.C. Beebe, Intracellular pH measurement using single excitation-dual emission fluorescence ratios, Ant. J. Physiol. 258: C171–C178 (1990)Google Scholar