Investigations into the use of handheld near-infrared spectrometer and novel semi-automated data analysis for the determination of protein content in different cultivars of Panicum miliaceumL.

A set of 42 millet (panicum miliaceum L.) samples was investigated for its protein content using standard Kjeldahl analysis and near-infrared spectroscopy. The performance of three handheld spectrometers was compared to a benchtop instrument. The used spectrometers operate in different regions of the NIR, which gives interesting insights into the applicability of each region. Additionally, semi-automated, consumer-oriented multivariate data analysis was compared to sophisticated data evaluation. The performance of the near-infrared instruments was compared using important statistical parameters of the established cross- and test set validated partial least squares regression (PLS-R) models. Milled and intact samples were analysed, in order to further evaluate the importance of homogeneity. The results showed that the benchtop spectrometer is capable of accurately analysing protein content of millet grains, with root mean square error (RMSEP) values for milled and intact grains of approximately 0.5%. Two PLS-R models of handheld instruments also yielded good results for milled grains with RMSEP values of about 0.6%. The semi-automated multivariate data analysis showed some drawbacks compared to standard data processing software. For intact grains, however, similar results could be achieved.