Some time ago, the food safety agency of the New Zealand Government recalled a brand of coconut milk powder due to the presence of an undeclared allergen – milk, in this case.
The recall is available at this link.
Inspired by this story, we decided to show how NIR analysis can be used to detect allergens. We mixed milk powder and coconut milk powder with different ratios, from 100% milk powder to 100% coconut milk powder in increments of 10%. We run NIR spectroscopy analysis on all samples to verify how the NIR spectra correlates to the different composition of the samples.
The first derivative spectra (average of 20 scans) of all samples are shown below.
The spectra can clearly detect the difference in composition, for instance in the wavelength region around 2200 nm the intensity of the signal grows with increasing milk powder concentration.
To be more quantitative, we ran Principal Component Analysis (PCA) on our data, in both 2D and 3D. In this case we used all spectra that were acquired, 20 spectra each sample, for 11 samples.
The 2D score plots of the first two PCA components are shown below for all samples. Most of the samples are very well clustered, meaning that they could be easily differentiated. Specifically, the presence of just a few percent of milk powder in coconut milk powder can be detected!
To increase the accuracy, we ran PCA using three components instead of two. This process enables an even finer discrimination between the different sample compositions. The results are shown in the 3D score plot below. Look at how the measurements that tend to overlap in 2D – for instance 30% and 40% milk content – are separated much better in 3D.
NIR analysis is an invaluable tool for food safety analysis and quality assurance. NIR analysis is non-destructive and can be implemented directly into the production line, for improved quality of foodstuff.
Would you like some help on using NIR spectrometry for your food analysis applications? Feel free to give us a buzz!