Skip to main content
ARS Home » Research » Publications at this Location » Publication #189704


item Kim, Yookyung
item Kays, Sandra

Submitted to: Pittsburgh Conference
Publication Type: Abstract Only
Publication Acceptance Date: 10/26/2005
Publication Date: 3/12/2006
Citation: Kim, Y., Kays, S.E. 2006. Simultaneous determination of major components and energy in homogenized meals using near-infrared spectroscopy [abstract]. Pittsburgh Conference. Paper No. 1250-4P.

Interpretive Summary:

Technical Abstract: One advantage of near-infrared spectroscopy (NIRS) is that it is a multi-analytical technique. Since the first study on analysis of protein content in soybeans, NIRS has proven to be an excellent method for nutrient analysis of foods. However, the use of NIRS in the analysis of compound or mixed meals is more complicated than for raw materials or one food group. Mixed meals contain two or more food groups and, thus, their spectra vary greatly as each raw material or ingredient has its own spectral pattern with virtually an infinite number of possible combinations. The objective of this study was to evaluate the feasibility of NIRS for prediction of moisture, crude protein, crude fat, total carbohydrate, ash, and energy simultaneously in homogenized mixed meals. Samples were selected from retailers and represented a wide range of frozen, packaged, and canned meals. Meals were homogenized and NIR spectra obtained with a dispersive NIR spectrometer. Major components were measured in dried samples by the appropriate AOAC method and reference values determined for homogenized samples. Total carbohydrate was calculated by difference and gross energy was estimated from the percentage crude protein, crude fat, and total carbohydrates using conversion factors of 4, 9, and 4 kcal/g, respectively. Using NIR reflectance spectra (1100 to 2496 nm) of homogenized samples and reference values for major components, partial least squares (PLS) regression models were developed for the prediction of major components (n=115). The PLS models were evaluated by cross validation and by prediction of an independent validation set (n=36). The R2 and SECV for PLS models for moisture, crude protein, crude fat, total carbohydrate, ash and gross energy were 0.98 and 1.74 (range 44.8 to 90.4) %; 0.98 and 1.11(range 1.9 to 23.9) per cent; 0.95 and 0.97(range 0.1 to 16.5) %; 0.96 and 2.09 (range 1.1 to 37.4) per cent; 0.92 and 0.24 (range 0.4 to 3.3) per cent; and 0.98 and 0.10 (range 0.4 to 2.94) kcal/g, respectively. The independent validation samples were predicted with an r2 of between 0.95 and 0.98 for all components except ash content. Overall performances of the models were found to be adequate for screening or quality control with RPD values between 3.1 and 5.7. The prediction accuracy was highest for moisture, followed by total carbohydrate, crude protein, crude fat and ash. NIRS provides an acceptable technique for the simultaneous analysis of moisture, protein, fat, carbohydrate, and energy in homogenized meals without need for further sample pretreatment.