Se. Kays et al., Predicting protein content by near infrared reflectance spectroscopy in diverse cereal food products, J NEAR IN S, 8(1), 2000, pp. 35-43
Simultaneous determination of constituents (e.g. dietary fibre, protein, fa
t) by near infrared (NIR) spectroscopy would increase the speed and efficie
ncy of nutrient analysis while substantially reducing the cost. Previous wo
rk has described the development of NIR reflectance models for the predicti
on of dietary fibre in a diverse group of cereal food products. While NIR s
pectroscopy has been used to measure protein content in cereal samples comp
rised of a single grain type, the utility of the NIR technique would be gre
atly improved if it could be expanded to cereal products derived from a div
erse cross-section of grains and formulations, The present study was conduc
ted to investigate the potential of NIR spectroscopy for the analysis of pr
otein in a data set that included products with numerous grains, such as wh
eat, oats, rice, rye, corn, millet, buckwheat and with a wide range of fat,
sugar and fibre contents. In addition, numerous processing techniques and
food additives were represented in the data set. Nitrogen content of dry-mi
lled cereal products was measured by combustion analysis (AOAC Method 992.2
3) and the range in nitrogen values was from 0.65 to 3.31% of dry weight. M
illed cereal products were scanned from 1100 to 2500nm with a scanning mono
chromator, A nitrogen calibration was developed, using a commercial analysi
s program, with modified partial least squares as the regression method. Th
e standard error of cross validation and R-2 for nitrogen (n = 147 calibrat
ion samples) were 0.090% and 0.973, respectively. Independent validation sa
mples (n = 72) were predicted with a standard error of performance of 0.079
% nitrogen and r(2) of 0.984. Because of the diversity of grains in the dat
a set, crude protein was calculated using two nitrogen-to-protein conversio
n methods and two PLS models were developed for the prediction of crude pro
tein. Crude protein was predicted with a similar precision to nitrogen and
the results for both protein models are within the precision required for U
S nutrition labelling legislation. In conclusion, NIR reflectance spectrosc
opy can be used for rapid and accurate prediction of nitrogen and crude pro
tein content in a heterogeneous group of cereal products comprised of a wid
e cross-section of grains and formulations.