Spectroscopy

Jens Petter Wold

Our activities within spectroscopy are applied, and focus on development of new rapid and non-destructive measurement applications for the food industry. We also implement on-line methods in manufacturing processes. To be able to find the best solution to a variety of problems, we master several spectroscopic techniques, which all can be applied more or less non-destructively on foods.

Content in this page:
1. Spectroscopy facts
2. Cooperation and development
3. Publications Spectroscopy (at Campus Ås)
4. Contacts and related links


1. Applied Food Spectroscopy

Our main methods are near-infrared spectroscopy (NIR), front face fluorescence spectroscopy, visible spectroscopy, FT-IR spectroscopy as well as Raman spectroscopy. Most of the work we do is on lab scale with laboratory instruments, however, we do also work with on-line instruments and miniature spectrometers in order to find dedicated and low-cost solutions when possible.
In all our work, multivariate statistics (chemometrics) is a fundamental tool either for explorative data analysis, regression, classification and experimental designs.

Solid sample fluorescence spectroscopy
We are working extensively on how autofluorescence from solid samples like meat and meat slurries can be used to quantify different constituents and quality parameters. Autofluorescence from biological tissues is often of low intensity demanding very sensitive optical detectors. The emission spectra can be quite complex, arising from a number of known, as well as unknown molecules, and the total fluorescence is sensitive to the local environment, apparently reducing the methods robustness. In spite of these difficulties and in the wake of constantly improving sensor technology and excellent optics more and more interesting applications within life sciences appear.
Traditional analysis of fluorescence spectroscopy is univariate, meaning that the intensity of fluorescence at one given excitation/emission wavelength pair is used to monitor an analyte of interest. The univariate analysis focuses on the variance of this single wavelength and does not take advantage of the information content of a complete spectrum. In short, such analysis of complex systems will be infested with noise and potentially useful data is thrown away. The problem of interference in fluorescence spectra can be overcome by using methods of multivariate analysis.
We have shown that fluorescence is an effective method to measure fat and connective tissue in meat. A specific area of interest is development of non-destructive measurements of lipid oxidation based on solid sample fluorescence. Promising results has been shown for poultry meat, complex meat products, dairy products and fish.

Raman spectroscopy
Raman spectroscopy has great potential for biochemical analysis of bio-tissue on both the macroscopic and microscopic scale. One major advantage of this technique is its ability to provide information about concentration, structure, and interaction of biochemical molecules within intact cells and tissues, nondestructively, without homogenization, extraction or the use of dyes or any other labeling agent. Raman spectroscopy has been recognized as a promising analytical tool within different areas of food science, and one such area is rapid and non-destructive quality assessment of foods for in-line purposes.
Raman scattering is a relatively weak optical effect that requires lasers for efficient excitation. In many biomaterials the fluorescence induced by this incident light is much more intense than the Raman scattered light, and this characteristic can make the Raman signals difficult to measure. The optimal wavelength region for Raman excitation for rapid tissue analysis is between 780 and 850 nm. Excitation in this region minimizes the fluorescence emission to an acceptable level and allows the use of sensitive CCD cameras which can capture high signal to noise spectra in short-time exposures (a few seconds). Several biomedical studies demonstrate that 785 and 850 nm Raman spectroscopy can be used for precise quantitative histochemical analysis of various types of human tissue.
Raman is a specific spectroscopic technique that measures rocking, wagging, scissoring, and stretching fundamental vibrations of functional groups containing bonds such as C-C, C-O, C-H and O-H. Some of the components contributing to the Raman scattering in muscle tissues are collagen, elastin, carotenoids, fatty acids, cholesterols, all constituents which can be useful to describe food quality. We have evaluated Raman with respect to food applications and have found it promising for measuring fat content and specific fatty acids, pigments, collagens, and secondary protein structures

Chemical imaging
Multispectral imaging is an interdisciplinary field comprising image analysis, spectroscopy and chemistry. The technique is well known from remote sensing of earth by satellite imagery, medical diagnostics and material science. Within food analysis there are a few applications, but the potential should be great. Compared to for instance earth remote sensing, we have the advantage of relatively easy access to the ground truth, represented as the distribution of biochemical properties which can be quantified in conventional laboratories. We do know the spectral properties of various prominent constituents in detail, and if not, these can in many cases be measured.

A multispectral image consists of several congruent images representing intensities at different wavelength bands. A multispectral image is then a stack of images where each pixel (meaning pixel vector or voxel) represents a spectrum for that specific point. Consequently, the image contains chemical information in every pixel. The main attractive feature of MI is the opportunity to create chemical maps where patterns and their kinetics can be studied in detail. For detailed food analysis, local regions of high analyte concentrations and concentration gradients are often more interesting than average concentrations, no matter how accurately the latter are determined. It is advantageous to know about and understand heterogeneity and the spatial propagation of chemical processes represented it in understandable images.

Multispectral images form huge data sets and requires powerful mathematical methods in order to get a sound overview of the images and spectral characteristics. We use the interactive approach of principal component analysis (PCA) to multivariate images, where the chemometric tradition of studying scores and loadings plots s actively used to explore the image data. This interactive work has the advantage of bringing an insight into the structure of the data set. Classes in the image can be studied and segmented via score plots, and the connection between spectral features and the raw image channels can be identified by investigating the loadings. When we have reference values for each pixel in the multi spectral images, image PLS regression has shown to be an excellent and efficient tool for modeling the images.
We have published examples where multispectral images are used to model distribution of lipid oxidation in poultry meat and light induced oxidation in dairy products by PLS. SIMCA classification is used in multispectral images in the VIS and NIR region to detect parasites in fish fillets.


2. Cooperation and development

We have world wide contacts within these scientific fields. In some projects we have close cooperation with other research institutes, mainly Norwegians; Norwegian Institute of Aquaculture and Fisheries Research (Fiskeriforskning) and Sintef. We do also have good contact with CPAC (Center for Process Analytical Chemistry) in Seattle, where the Raman studies so far were done.

Updated: 18.11.02
Printer-friendly version : Archive : Notify a friend