Classification before regression for improving the accuracy of glucose quantification using absorption spectroscopy.

This work contributes to the improvement of glucose quantification using near-infrared (NIR), mid-infrared (MIR), and combination of NIR and MIR absorbance spectroscopy by classifying the spectral data prior to the application of regression models. Both manual and automated classification are presented based on three homogeneous classes defined following the clinical definition of the glycaemic ranges (hypoglycaemia, euglycaemia, and hyperglycaemia). For the manual classification, partial least squares and principal component regressions are applied to each class separately and shown to lead to improved quantification results compared to when applying the same regression models for the whole dataset. For the automatic classification, linear discriminant analysis coupled with principal component analysis is deployed, and regressions are applied to each class separately. The results obtained are shown to outperform those of regressions for the entire dataset.

Coupling Scatter Correction with bandpass filtering for preprocessing in the quantitative analysis of glucose from near infrared spectra.

This paper proposes a novel pre-processing method based on combining bandpass filtering with scatter correction techniques Multiplicative Scatter Correction (MSC) and Standard Normal Variate (SNV) to enhance the prediction capability of the linear regression models Partial Least Squares Regression (PLSR) and Principal Component Regression (PCR) in near infrared (NIR) spectroscopy. The method is implemented into a calibration model, evaluated and then validated for the prediction of the glucose concentration from NIR spectra of an aqueous mixture of human serum albumin and glucose in a solution of distilled water and phosphate buffer. The results obtained demonstrate improved prediction performance for both PCR and PLSR. Compared to the efficient feature weighting pre-processing (RRelief), the proposed method is shown to yield better prediction reducing the Root Mean Square Error Prediction RMSEP.

Savitzky-Golay coupled with digital bandpass filtering as a pre-processing technique in the quantitative analysis of glucose from near infrared spectra.

This paper proposes a novel pre-processing method based on combining bandpass with Savitzky-Golay filtering to further improve the prediction performance of the linear calibration models Principal Component Regression (PCR) and Partial Least Squares Regression (PLSR) in near infrared spectroscopy. The proposed method is compared to the highly efficient RReliefF pre-processing technique for further evaluation. The developed calibration models have been validated to predict the glucose concentration from near infrared spectra of a mixture of glucose and human serum albumin in a phosphate buffer solution. The results show that the proposed technique improves the prediction performance of both the PCR and PLSR models and achieve better results than the RReliefF technique.