Application of a Handheld Near-Infrared Spectrometer to Predict Gelatinized Starch, Fiber Fractions, and Mineral Content of Ground and Intact Extruded Dry Dog Food

SIMPLE SUMMARY: The pet food industry is interested in performing fast analyses to control the nutritional quality of their products. Despite having some limitations related to the need to modify the production process or to have a laboratory to prepare the samples for analysis through desktop instruments, near-infrared spectroscopy is one of the most used technologies for inexpensive analysis of foodstuffs. Thus, the miniaturization of infrared devices allows a wider industrial applicability of this technique. Information on the use of miniaturized infrared tools in the pet food sector is currently very limited, and the present research is the first attempt to predict the total and gelatinized starch, insoluble fibrous fractions, and mineral content of ground and intact dry pet food using the handheld NIR scanner SCiO™. Results from the current study revealed no significant differences in the predictive ability of the instrument using both ground and intact samples. The instrument offers a potential for screening purposes of both total and gelatinized starch, revealing the potential to monitor their content and ratio in commercial dog food on a large scale. Improvements such as widening the wavelength range is expected to increase prediction models’ accuracy. ABSTRACT: The aim of the present study was to investigate the ability of a handheld near-infrared spectrometer to predict total and gelatinized starch, insoluble fibrous fractions, and mineral content in extruded dry dog food. Intact and ground samples were compared to determine if the homogenization could improve the prediction performance of the instrument. Reference analyses were performed on 81 samples for starch and 99 for neutral detergent fiber (NDF), acid detergent fiber (ADF), acid detergent lignin (ADL), and minerals, and reflectance infrared spectra (740 to 1070 nm) were recorded with a SCiO™ near-infrared (NIR) spectrometer. Prediction models were developed using modified partial least squares regression and both internal (leave-one-out cross-validation) and external validation. The best prediction models in cross-validation using ground samples were obtained for gelatinized starch (residual predictive deviation, RPD = 2.54) and total starch (RPD = 2.33), and S (RPD = 1.92), while the best using intact samples were obtained for gelatinized starch (RPD = 2.45), total starch (RPD = 2.08), and K (RPD = 1.98). Through external validation, the best statistics were obtained for gelatinized starch, with an RPD of 2.55 and 2.03 in ground and intact samples, respectively. Overall, there was no difference in prediction models accuracy using ground or intact samples. In conclusion, the miniaturized NIR instrument offers the potential for screening purposes only for total and gelatinized starch, S, and K, whereas the results do not support its applicability for the other traits.