Authentication of edible oils using Fourier transform infrared spectroscopy and pattern recognition methods

Abstract

Fourier transform infrared (FTIR) spectra of ninety-seven edible oil samples from twenty plant based varieties collected over a three year period were analyzed using the four main types of pattern recognition methods: mapping and display, cluster analysis, feature selection, and classification. The ninety-seven edible oil samples selected for this study encompassed multiple brands representing supplier to supplier variation as well as both seasonal and batch variation within a supplier. Using a hierarchical classification scheme, the twenty plant based varieties of edible oils could be divided into four distinct groups. Edible oils from different oil groups could be reliably discriminated, whereas the discrimination of edible oils within the same group was shown to be problematic. Adulteration of the plant based edible oils by other edible oils in the same group (e.g., EVOO by almond oil) could not be reliably detected using FTIR spectroscopy, whereas adulteration of edible oils by other oils that were not in the same group (e.g., EVOO adulterated by corn or canola oil) could be detected at concentration levels as low as 10% (v/v) which is consistent with the results reported in previously published studies using partial least squares regression. A unique aspect of this work is the incorporation of edible oils collected systematically over three years, which introduced a heretofore unseen variability in the chemical composition of the edible oils. This work also demonstrates that previously published studies (which have relied on a single source to represent each type of edible oil) provide an overly optimistic estimate of the capability of FTIR spectroscopy to discriminate plant based edible oils by type and to detect the presence of adulterants in plant based edible oils.