MULTIVARIATE CALIBRATION FOR PHYSIOLOGICAL SAMPLES USING INFRARED-SPECTRA WITH CHOICE OF DIFFERENT INTENSITY DATA

A multicomponent assay several blood substrates is presented for human plasma using mid-infrared spectra recorded by the attenuated total re flection technique. Two different sample populations were analysed: a pool plasma spiked with different amounts of glucose and a hospital po pulation of samples from 126 different patients. Partial least-squares was used for multivariate calibration based on spectral intervals in the fingerprint region selected for optimum prediction modeling. Diffe rent data, either plasma absorbance spectra with compensation of the w ater signal or single beam spectra, were considered. Results are repor ted for protein, glucose, cholesterol, triglycerides and urea.