Direct determination of major components in human diets and baby foods.

Abstract

A fast method has been developed for the determination of fat, proteins, carbohydrates, and energy value in baby food, infant fast food, and canteen menus, using near-infrared (NIR) and mid-infrared (MIR) spectroscopy measurements and multivariate calibration methods based on partial least square regression. Reference standard methods were employed to build and validate the infrared methods for direct determination of nutrients. Coefficients of determination obtained between predicted values and reference ones for total fat, proteins, carbohydrates, and energy value were 96.7, 98.1, 98.9, and 96.5 for NIR and 91.0, 93.0, 92.0, and 84.1 for MIR, respectively, with relative root mean square error of prediction (RRMSEP) below or equal to 9% for NIR and 16% for MIR. Results obtained indicate that both NIR and MIR techniques have good predictive capabilities, with the NIR method being the most accurate and simple.