RICE QUALITY AND PROCESSING Apparent Amylose Content Prediction using Near Infrared Spectroscopy of Individual and Bulk Rice Kernels

For this study seventy-eight rice samples from 2008 of mediumand long-grain rice cultivars were harvested and selected from Rohwer, Keiser, and Corning locations in Arkansas. Bulk milled rice samples of ~11.5 grams and 100 individual milled rice kernels fitted with single kernel adapter were scanned using a diode-array-analyzer of each cultivar/location. Partial least squares regression (PLS) was performed using Unscrambler software to develop prediction models of bulk and individual rice amylose content. Measured and predicted bulk and single kernels amylose content recorded coefficient of determinations of 0.83 for bulk kernels and 0.80 for the single kernel using PLS, respectively. The correspondent root mean square errors of prediction (RMSEP) were 2.33 for bulk amylose measurements and 2.55 for single kernel measurement. Distributions of amylose content of individual and bulk kernels were evaluated as indications of wet chemistry amylose analysis. Amylose content of individual and bulk kernels was normally distributed around means that are correlated with amylose content of milled rice bulk and standard wet chemistry amylose analysis.