Skip to main content
ARS Home » Research » Publications at this Location » Publication #140865
Title: DEVELOPMENT OF ROBUST CALIBRATION MODELS FOR PROTEIN AND AMYLOSE IN RICE FLOUR USING FT-RAMAN AND NIR SPECTROSCOPY

Author
item SOHN, MI RYEONG - USDA-FAS
item Himmelsbach, David
item Barton Ii, Franklin

Submitted to: Eastern Analytical Symposium
Publication Type: Abstract Only
Publication Acceptance Date: 4/1/2002
Publication Date: 11/1/2002
Citation: SOHN, M.R., HIMMELSBACH, D.S., BARTON II, F.E. DEVELOPMENT OF ROBUST CALIBRATION MODELS FOR PROTEIN AND AMYLOSE IN RICE FLOUR USING FT-RAMAN AND NIR SPECTROSCOPY. THE 41ST EASTERN ANALYTICAL SYMPOSIUM. 2002. ABSTRACT NO. 435. P. 40.

Interpretive Summary:

Technical Abstract: Spectroscopic analysis has been investigated with the aim of developing a database for the evaluation of rice quality. The purpose of this study is to develop robust Near infrared Fourier transform Raman (NIR-FT/Raman) and Near infrared (NIR) spectroscopy calibration models for determining the protein and amylose content in rice flour, which reduce the effect between samples having different cultivation year. A total of 222 rice samples, 90 from 1996 and 132 from 1999, collected in US, Taiwan, Korea and Australia were used in the calibration model. The spectra of milled rice flour samples were taken in range of 3600-230 cm**-1 Raman regions with a 1064 nm NIR excitation laser at 500mW of power. NIR spectra were taken in the range of 1100-2500 nm regions using a spinning cup. NIR and FT-Raman spectral data were preprocessed by background correction due to changing the instrument parts. Spectral data were then processed with baseline correction and normalization for FT-Raman and multiplicative scatter correction for NIR. Smoothing, mean normalization, 1st or 2nd derivative, and Marten's Uncertainty were used to improve calibration models. All PLS models were carried out using Unscrambler software. Robust calibrations were made by combination of 2 sample sets and by using spectral pre-treatments. Calibration models showed better results after Marten's Uncertainty with both spectroscopic techniques. FT-Raman calibration developed for protein was the best in this sample set with r**2=0.941, SEP=0.4%, and Bias=0.000% using 4 factors and smoothing processing. Amylose calibration was best with r**2=0.956, SEP=1.572%, and Bias=-0.006% using 8 factors and no pre-treatment. PLS models for both constituents resulted in better calibration in NIR than in FT-Raman. 1st derivative processing improved the models for the two constituents in NIR. Smoothing, mean normalization and 2nd derivative did not improve results. The best protein model by NIR was obtained by using 2 factors and 1st Norris derivative processing giving r**2=0.973, SEP=0.282%, and Bias=-0.000%. The best amylose model by NIR was obtained using 8 factors and 1st Savitsky-Golay derivative processing giving r**2=0.976, SEP=1.155%, and Bias=-0.008%.