Vegard Segtnan

Raman Spectra of Biological Samples: A Study of Preprocessing Methods

In this study preprocessing of Raman spectra of different biological samples has been studied, and their effect on the ability to extract robust and quantitative information has been evaluated. Four data sets of Raman spectra were chosen in order to cover different aspects of biological Raman spectra, and the samples constituted salmon oils, juice samples, salmon meat, and mixtures of fat, protein, and water. A range of frequently used preprocessing methods, as well as combinations of different methods, was evaluated. Different aspects of regression results obtained from partial least squares regression (PLSR) were used as indicators for comparing the effect of different preprocessing methods. The results, as expected, suggest that baseline correction methods should be performed in advance of normalization methods. By performing total intensity normalization after adequate baseline correction, robust calibration models were obtained for all data sets. Combination methods like standard normal variate (SNV), multiplicative signal correction (MSC), and extended multiplicative signal correction (EMSC) in their basic form were not able to handle the baseline features present in several of the data sets, and these methods thus provide no additional benefits compared to the approach of baseline correction in advance of total intensity normalization. EMSC provides additional possibilities that require further investigation.

Non-contact transflectance near infrared imaging for representative on-line sampling of dried salted coalfish (bacalao)

This paper describes a multi-spectral imaging near infrared (NIR) transflectance system developed for on-line determination of crude chemical composition of highly heterogeneous foods and other bio-materials. The system was evaluated for moisture determination in 70 dried salted coalfish (bacalao), an extremely heterogeneous product. A spectral image cube was obtained for each fish and different sub-sampling approaches for spectral extraction and partial least squares calibration were evaluated. The best prediction models obtained correlation R2 values around 0.92 and root mean square error of cross-validation of 0.70%, which is much more accurate than todays traditional manual grading. The combination of non-contact NIR transflectance measurements with spectral imaging allows rather deep penetrating optical sampling as well as large flexibility in spatial sampling patterns and calibration approaches. The technique works well for moisture determination in heterogeneous foods and should, in principle, work for other NIR absorbing compounds such as fat and protein. A part of this study compares the principles of reflectance, contact transflectance and non-contact transflectance with regard to water determination in a set of 20 well-defined dried salted cod samples. Transflectance and non-contact transflectance performed equally well and were superior to reflectance measurements, since the measured light penetrated deeper into the sample.

Noncontact salt and fat distributional analysis in salted and smoked salmon fillets using X-ray computed tomography and NIR interactance imaging.

To be able to monitor the salting process of cold smoked salmon, a nondestructive imaging technique for salt analysis is required. This experiment showed that X-ray computed tomography (CT) can be used for nondestructive distributional analysis of NaCl in salmon fillets during salting, salt equilibration, and smoking. The combination of three X-ray voltages (80, 110, and 130 kV) gave the best CT calibrations for NaCl, with a prediction error (root mean square error of cross-validation, RMSECV) of 0.40% NaCl and a correlation (R) of 0.92 between predicted values and reference values. Adding fat predictions based on NIR interactance imaging further improved the NaCl prediction performance, giving RMSECV = 0.34% NaCl and R = 0.95. It was also found that NIR interactance imaging alone was able to predict NaCl contents locally in salted salmon fillets with RMSECV = 0.56% and R = 0.86.

Raman and Near-Infrared Spectroscopy for Quantification of Fat Composition in a Complex Food Model System

Raman and near-infrared (NIR) spectroscopy have been evaluated for determining fatty acid composition and contents of main constituents in a complex food model system. A model system consisting of 70 different mixtures of protein, water, and oil blends was developed in order to create a rough chemical imitation of typical fish and meat samples, showing variation both in fatty acid composition and in contents of main constituents. The model samples as well as the pure oil mixtures were measured using Raman and NIR techniques. Partial least squares regression was utilized for prediction, and fatty acid features were expressed in terms of the iodine value and as contents of saturated, monounsaturated, and polyunsaturated fatty acids. Raman spectroscopy provided the best results for predicting iodine values of the model samples, giving validated estimation errors accounting for 2.8% of the total iodine value range. Both techniques provided good results for predicting the content of saturated, monounsaturated, and polyunsaturated fatty acids in the model samples, yielding validated estimation errors in the range of 2.4–6.1% of the total range of fatty acid content. Prediction results for determining fatty acid features of the pure oil mixtures were similar for the two techniques. NIR was clearly the best technique for modeling content of main constituents in the model samples.

Fat distribution analysis in salmon fillets using non-contact near infrared interactance imaging: a sampling and calibration strategy

An online NIR interactance imaging instrument was tested for fat distribution analysis in raw and salted salmon fillets. Approximately 3000 spectra were collected for each fillet when passing under the instrument on a conveyor belt (approximately 1s exposure). The instrument was calibrated using five cylindrical plugs (15 mm diameter) from each fillet. The fat content was measured for each of these plugs using 1H-NMR spectrometry and the spectra from each plug region were averaged and used for calibration and validation. It was found that online NIR interactance imaging is well-suited for distributional fat analysis in raw and salted intact salmon fillets. The local sampling and calibration strategy using 15 mm diameter plugs for reference analysis and spectral averaging was found to provide relevant information and robust models. The average prediction errors (root mean square error of cross-validation) for raw and salted fillets in combination were approximately 2% fat for local plug regions.

Temperature, sample and time dependent structural characteristics of gelatine gels studied by near infrared spectroscopy

Abstract The aim of the present experiment was to investigate differences in the NIR spectra of gelatine gels and to relate spectral differences to structural characteristics. Spectral changes induced by temperature, maturation time and sample-related differences, like the Bloom strength, viscosity and sample origin, have been studied. The main conclusion is that the same or very similar spectral changes take place in gelatine gels at decreasing temperatures, increasing Bloom strengths and increasing maturation time, and that the changes are most clearly seen in the 2160–2210 nm region. It is suggested that the 2160–2170 nm region represents single chains (with or without helical parts), while bands 2180 and 2208 nm represent triple helices and β-turns. In the region containing the first OH overtone of water, increased intensities were observed with increased maturation time at the 1414 and 1462 nm bands, along with a decrease at the 1492 nm band. No temperature- or sample-dependent spectral changes were observed in the 1400–1500 nm region that cannot be explained by temperature changes in bulk water. From the NIR spectra it was possible to distinguish hide gelatine gels from pigskin gelatine gels. The overall absorbance level was found to increase slightly with time, indicating that some aggregation takes place as a function of gel maturation time.

Low-Cost Approaches to Robust Temperature Compensation in Near-Infrared Calibration and Prediction Situations

The traditional way of handling temperature shifts and other perturbations in calibration situations is to incorporate the non-relevant spectral variation in the calibration set by measuring the samples at various conditions. The present paper proposes two low-cost approaches based on simulation and prior knowledge about the perturbations, and these are compared to traditional methods. The first approach is based on augmentation of the calibration matrix through adding simulated noise on the spectra. The second approach is a correction method that removes the non-relevant variation from new spectra. Neither method demands exact knowledge of the perturbation levels. Using the augmentation method it was found that a few, in this case four, selected samples run under different conditions gave approximately the same robustness as running all the calibration samples under different conditions. For the carbohydrate data set, all robustification methods investigated worked well, including the use of pure water spectra for temperature compensation. For the more complex meat data set, only the augmentation method gave comparable results to the full global model.

Rapid assessment of physico-chemical properties of gelatine using near infrared spectroscopy

Abstract The aim of the present study was to investigate the feasibility of near infrared (NIR) spectroscopy for rapid assessment of the Bloom value, viscosity, pH and moisture content of commercial edible gelatines. NIR spectra from both dry gelatine samples and freshly made gels (matured for 30 min) were calibrated against the four parameters. The lowest prediction errors for Bloom value and pH were obtained from the gel spectra, while dry gelatine spectra provided the lowest prediction errors for viscosity and moisture content. However, all four parameters could be predicted from dry gelatine spectra with cross-validated correlation coefficients between 0.80 and 0.86. All calibration models gave correlation coefficients in the region 0.80–0.92.

Near-Infrared Hyperspectral Imaging of Fusarium-Damaged Oats (Avena sativa L.)

ABSTRACT The feasibility of hyperspectral imaging (HSI) to detect deoxynivalenol (DON) content and Fusarium damage in single oat kernels was investigated. Hyperspectral images of oat kernels from a Fusarium-inoculated nursery were used after visual classification as asymptomatic, mildly damaged, and severely damaged. Uninoculated kernels were included as controls. The average spectrum from each kernel was paired with the reference DON value for the same kernel, and a calibration model was fitted by partial least squares regression (PLSR). To correct for the skewed distribution of DON values and avoid nonlinearities in the model, the DON values were transformed as DON* = [log(DON)]3. The model was optimized by cross-validation, and its prediction performance was validated by predicting DON* values for a separate set of validation kernels. The PLSR model and linear discriminant analysis classification were further used on single-pixel spectra to investigate the spatial distribution of infection in the kerne...

Incorporating interactions in multi-block sequential and orthogonalised partial least squares regression

This paper is about how to incorporate interaction effects in multi‐block methodologies. The method proposed is inspired by polynomial regression modelling in the case with only a few independent variables but extends/generalises the idea to situations where the blocks are potentially very large with respect to the number of variables. The method follows a so‐called type I sums of squares strategy where the linear effects (main effects) are incorporated sequentially and before the interactions. The sequential and orthogonalised partial least squares (SO‐PLS) technique is used as a basis for the proposal. The SO‐PLS method is based on sequential estimation of each new block by the PLS regression method after orthogonalisation with respect to blocks already fitted. The new method preserves the invariance already established for SO‐PLS and can be used for blocks with different dimensionality. The method is tested on one real data set with two independent blocks with different complexity and on a simulated data set with a large number of variables in each block. Copyright