MULTIVARIATE DETERMINATION OF GLUCOSE IN WHOLE-BLOOD USING PARTIAL LEAST-SQUARES AND ARTIFICIAL NEURAL NETWORKS BASED ON MIDINFRARED SPECTROSCOPY

The infrared (IR) spectra of whole blood EDTA samples, in the range be tween 1500 and 750 cm-1, obtained from the patient population of a gen eral hospital, were used to compare different multivariate calibration techniques for quantitative glucose determination. Ninety-six spectra of whole undiluted blood samples with glucose concentration ranging b etween 44 and 291 mg/dL were used to create calibration models based o n a combination of partial least-squares (PLS) and artificial neural n etwork (ANN) methods. The prediction capabilities of these calibration models were evaluated by comparing their standard errors of predictio n (SEP) with those obtained with the use of PLS and principal componen t regression (PCR) calibration models in an independent prediction set consisting of 31 blood samples. The optimal model based on the combin ed PLS-ANN produced smaller SEP values (15.6 mg/dL) compared with thos e produced with the use of either PLS (21.5 mg/dL) or PCR (24.0 mg/dL) methods. Our results revealed that the combined PLS-ANN models can be tter approximate the deviations from linearity in the relationship bet ween spectral data and concentration, compared with either PLS or PCR models.