In this project, we will determine the most accurate technique for extracting concentrations of analytes from Raman spectra of blood samples. Although we have been successful in using partial least squares (PLS) to analyze raw spectral data, we may be able to improve our calibration accuracy by preprocessing the data to remove known sources of background variation or to account for nonlinearities in the system. In particular, the effects of variable sample turbidity upon signal strength will be studied. In addition, algorithms from the literature for selecting the optimal region of the spectrum for calibration will be studied.
Showing the most recent 10 out of 178 publications