Izzo, Christopher G.. Near infrared spectroscopic investigation of lipid oxidation in model solid food systems. Retrieved from https://doi.org/doi:10.7282/t3-danm-0315
DescriptionReports of the application of near infrared (NIR) spectroscopy to analyses of lipid oxidation in solid foods generally indicate poor performance. To elucidate reasons for this, effects of sample packing and presentation (off-centered rotation) on NIR analyses were examined in a sampling system miniaturized to employ amounts of material feasible for research studies. Packing and presentation conditions affording the best performance in qualitative studies were utilized in quantitative assays to determine the ability of NIR to monitor lipid oxidation in model solid food systems by comparison with reference chemical analyses of conjugated dienes, lipid hydroperoxides, and carbonyl products.
Preliminary investigation indicated constant forming pressure and rotational averaging during scanning reduced variation among replicate scans of mixtures of up to 15% (w:w) lipid with white rice flour. Neat pecan or canola oils oxidized at 40°C for up to sixteen weeks and assayed chemically for conjugated dienes, lipid hydroperoxides and carbonyls were used to prepare 7.5% (w/w) oil : white rice flour samples for NIR analysis with constant pressure and rotation. Canola oxidized more readily than pecan oil, reaching apparent maxima for conjugated dienes and peroxides; however, carbonyls developed only near the end of incubation.
NIR models of oxidation used either the full spectrum (4000 - 10,000 cm-1) or wavenumber ranges selected by statistical model improvement techniques. Full spectrum models of conjugated dienes or peroxides for pecan oil samples showed very poor correlations with chemical analyses; neither was improved by wavenumber selection. Full spectrum models for canola oil samples were slightly better and improved with wavenumber selection.
Peroxide value model quality rose with sample numbers; the opposite occurred for conjugated dienes. The best peroxide value models included far fewer wavenumbers than conjugated diene models, which were more susceptible to interference from various sources. Results from wavenumber selection appeared pathway dependent, varying with samples used and pretreatments applied in the initial model. Spectral reproducibility among nominally identical samples was the primary hindrance to quantitative correlations for conjugated dienes and peroxide values. Thus, improvements in sample presentation mechanisms and software may render NIR suitable for quantitative analysis of lipid oxidation in solid food systems.