Dr. Dieter Bingemann, Applications Scientist, Ocean Optics, Ostfildern,Germany, discusses the recent increase in the application breadth of near-infrared spectroscopy.
Near-infrared spectroscopy (NIR), nestled between the well-established analytical workhorses of visible and mid-infrared spectroscopy, has recently seen a dramatic increase in its application breadth for the food industry. While the results are not as straightforward to interpret or analyse as its two neighbours on either side on the wavelength scale, modern chemometric tools and smaller, more affordable NIR equipment have opened up near-infrared’s treasure trove of spectroscopic information and made it more easily used than ever before.
With NIR, practitioners can now combine the benefits of no sample preparation and rapid non-contact measurements with the specificity inherent in all vibrational spectroscopy techniques. With smaller, more affordable equipment, NIR can be applied to a greater range of applications than ever before, including the determination of many parameters in difficult to measure opaque and inhomogeneous foodstuffs. In this article, we explore the application of a small NIR spectrometer with modular light source and fiber optic probe in the development of a robust model for the non-contact, non-destructive measurement of apple sweetness using the Brix scale.
NIR spectroscopy – a vibrational technique
Near-infrared spectroscopy probes the vibrational overtone absorption of chemical bonds, predominantly CH, OH, and NH, and is therefore universally sensitive to almost any organic or edible sample. The resulting spectra are more complex than they appear, a single broad peak can contain multiple overlapping narrow absorption bands from different bonds in the same molecule or from different molecules in the same sample. Chemometrics, a powerful computational tool set, enables the interpretation of nearinfrared spectra and provides qualitative and quantitative answers in all areas of scientific measurements. Typical applications are the quantification of a concentration, the identification of a sample and the test against a quality standard.
The cutoff wavelength of the near-infrared detectors used in NIR spectrometers can be adjusted through the ratio of indium to gallium in the utilised InGaAs semiconductor material. Shorter cutoff wavelengths reduce noise and dark current, such that photodiode arrays for the wavelength range from 0.9 to 1.7 µm can even be operated without thermoelectric cooling, simplifying design and power requirements.
Predicting apple sweetness
The sugar content of fruit (primarily fructose, glucose, and sucrose) is commonly determined in the expressed fruit juice in sum as the soluble solids content (SSC). It is measured with a refractometer and is reported as degrees Brix (°Bx), or g sucrose equivalent / 100mL. Typical values for apples range from 10°Bx to 16°Bx, depending on the variety, with unripe and ripe apples of the same variety differing by up to 4°Bx. A Brix measurement is timeconsuming and destroys the fruit; near-infrared reflectance spectroscopy using chemometric analytical methods offers a nondestructive alternative.
The experimental setup combined a small modular NIR spectrometer with a halogen light source and a fiber optic reflectance probe. This is shown inset in Figure 1. The reflectance spectra for 76 Ginger Gold variety apples, both ripe and unripe, were measured and compared against a Brix value determined in a separate lab analysis. Spectra from 5 locations across each apple’s ‘equator’ were averaged to find representative spectra for the entire apple. An example is shown in Figure 1; due to slight differences in the measurement geometry and the shape of the apple, the raw spectra appear shifted and scaled relative to each other. This was compensated for using a pre-processing step applying the standard normal variate method. Finally, we subtracted the average of all spectra as only the differences between the spectra contain the information about the varying sweetness. The actual Brix value is then determined using a Partial Least Square (PLS) Regression approach.
In order to develop a robust model it is important to have both a test and a training set. The samples were randomly split into 1/3 for the test set and 2/3 for the training set. Cross-validation helps to walk the fine line between poor prediction because of a lack of data (underfitting) and poor prediction for future unknown samples despite good performance on the training set (overfitting). In this study the best model performance is achieved by including 5 components (basis vectors or dimensions) in the PLS model. The quality of the prediction is shown in Figure 2, which compares the
apple’s sweetness predicted by the model from the NIR reflectance spectrum for each apple with the actual Brix value measured in the lab for apples from both the training and test set. The
deviation between predicted and actual values is summarised in the ‘standard error of prediction’ (SEP), a measure of the quality of the model, which is better than 0.3°Bx in this investigation.
Near-infrared spectroscopy applications for food
Besides the apple sweetness measurement illustrated above, numerous parameters, such as the amount of moisture protein, starch, sugars, fat, to name but a few, can be determined in food using NIR techniques. Near-infrared spectroscopy can also detect unwanted ingredients in food, such as lactose in lactose-free milk, gluten in gluten-free grains, and even toxic compounds, such as glycol added to wine to artificially increase the apparent sweetness, or toxic melamine in milk powder to falsely create higher protein content.
All of these applications benefit from the universality of the spectral detection, the ease of sample preparation and the ability of NIR spectroscopy to provide reliable qualitative and quantitative answers quickly and affordably. With equipment that is increasingly smaller, easier to use and more affordable the application of NIR spectroscopy and chemometrics in the food industry has a bright future ahead.