In case of spices or crude drugs of medicinal- and aromatic plant origin, sensory characteristics, especially odour, has great commercial importance. The instrumental sensory analysis the so-called 'electronic nose' has proved to be a significant, new and quick method in chemometry. The sensor signal responses (data recorded by the electronic nose instrument) of the electronic nose were evaluated by statistical methods, including principal component analysis (PCA) and canonical discriminant analysis (CDA) and the combination of these methods by applying the discriminant analysis on the first eight principal components. The aim of this paper is the comparative analysis of the above evaluation methods as data processing tools of the sensor signal response of the electronic nose (chemosensor array). The essential oil of oregano (Origanum vulgare subsp. hirtum) selected line No. 10 was compared to the oil distilled from the selected line No. 11; and dried root samples of lovage (Levisticum officinale) harvested at different times from the two- and three-year-old population, were investigated with electronic nose (NST-3320, AppliedSensor Sweden AB). Principal component analysis, as a first step of the evaluation, did not clearly distinguish either oregano or lovage samples. Further statistical evaluation of the original sensor signal responses of the electronic nose with canonical discriminant analysis improved the separation power of the model. The best separation could be achieved by the combination of the two methods, whereby canonical discriminant analysis was applied to the first eight principal components, which described 99% of the differences. In all cases more than 92%, while in several experiments 100% of cross-validated grouped cases were classified correctly. Based on the results, the application of the electronic nose and the combination of multivariate methods, PCA and CDA, could be an appropriate tool either for identification of cultivar to accelerate selection process or to distinguish crude drugs of different age or different harvesting period.
The near infrared spectra are useful information sources relating to quality (e.g. composition) of a material examined. To obtain and interpret useful information requires in most cases the application of sophisticated methods of mathematical statistics. A method different from those mentioned above, implementing a large scale data reduction based on geometrical consideration is the PQS. According to this method, the quality of a material can be characterised by the centre of its spectrum represented in a polar co-ordinate system. In many cases it is enough to know whether the investigated product deviates in a certain degree from a given “standard product” or not. This can be decided by determining special “distances” between the two (investigated and standard) products using their near infrared spectra. Besides the successfully used Euclidean and Mahalanobis distances a new one, the “polar distance” was introduced giving the distance between the two centres (quality points) of the spectra of the two products examined. A method was elaborated to select the optimal wavelength range giving the maximum normalised distance between the two quality points of the investigated products. The so called “wavelength range optimisation” can not be used to work with non spectral data sets. While in case of NIR spectra the sequence of the data are determined by nature, in several cases the order of the data can be freely varied and the goal is the determination of the optimal data sequence. By introducing the “sequence optimisation” PQS could be generalised and used from the field of near infrared spectroscopy to solve any kind of multivariate tasks. The advantage of the PQS optimisation method is its simplicity. Since PQS was developed to extract the needed information from NIR spectra, the basic principles of the technique are introduced with the help of near infrared spectra of some milk powder samples of different fat content. The sequence optimisation is demonstrated with the sensor signal responses of an electronic nose (chemosensor array) instrument measuring different steam distilled volatile oil samples.
Authors:Kinga Horváth, Zs. Seregély, I. Dalmadi, Éva Andrássy, and J. Farkas
The utility of chemosensor array (EN) signals of head-space volatiles of aerobically stored pork cutlets as a non-invasive technique for monitoring their microbiological load was studied during storage at 4, 8 and 12 °C, respectively. The bacteriological quality of the meat samples was determined by standard total aerobic plate counts (TAPC) and colony count of selectively estimated
(PS) spp., the predominant aerobic spoilage bacteria. Statistical analysis of the electronic nose measurements were principal component analysis (PCA), and canonical discriminant analysis (CDA). Partial least squares (PLS) regression was used to model correlation between microbial loads and EN signal responses, the degree of bacteriological spoilage, independently of the temperature of the refrigerated storage. Sensor selection techniques were applied to reduce the dimensionality and more robust calibration models were computed by determining few individual sensors having the smallest cross correlations and highest correlations with the reference data. Correlations between the predicted and “real” values were given on cross-validated data from both data reduced models and for full calibrations using the 23 sensor elements. At the same time, sensorial quality of the raw cutlets was noted subjectively on faultiness of the odour and colour, and drip formation of the samples. These preliminary studies indicated that the electronic nose technique has a potential to detect bacteriological spoilage earlier or at the same time as olfactory quality deterioration.
Authors:K. Horváth, Zs. Seregély, É. Andrássy, I. Dalmadi, and J. Farkas
Experimental batches of chilled boneless slices of pork meat have been stored aerobically in sterile Petri dishes and total aerobic plate counts (TAPC) and sensorial observations were made periodically during storage to monitor bacterial growth and apparent deteriorative changes at 4, 8 and 12 °C, respectively. Near infrared spectroscopy (diffuse reflectance) measurement was performed on replicate meat samples in the wavelength range of 1000–1800 nm. Second derivative and multiplicative scatter correction were performed on the spectra as data pre-treatments. Principal component analysis (PCA) and canonical discriminant analysis (CDA) were used for observation of discrimination of the samples due to loss of freshness and onset of bacterial spoilage as a function of the storage time. The percentage of correctly classified samples decreased somewhat by increasing the storage temperature. Partial least squares (PLS) chemometric model was developed to predict and quantify bacterial loads from the scatter corrected 2nd derivative spectra. PLS evaluation (predicted versus measured TAPC values) — when bacterial counts at all sampling days and storage temperatures were taken into account — resulted in a correlation coefficient of 0.977, and a root mean square error of prediction (RMSEP) 0.438 log colony forming units/g. These preliminary results indicate the potential of utilising near infrared diffuse reflectance spectroscopy in combination with multivariate statistical methods to monitor loss of freshness and detect bacterial spoilage of meat samples rapidly before deleterious microbial changes become apparent. However, much larger number of samples should be studied to ascertain properly the prediction power of the spectroscopic method.
Authors:É. Andrássy, J. Farkas, Zs. Seregély, I. Dalmadi, E. Tuboly, and V. Lebovics
Experiments were performed to study changes caused by irradiation or high hydrostatic pressure pasteurization of liquid egg white by differential scanning calorimetry, spectrofluorimetry, electronic nose measurements and NIR-spectrometry. The non-thermal pasteurization treatments were also assessed in relation to loss of carotenoid content, and lipid- and cholesterol oxidation of liquid egg yolk. Unlike radiation pasteurization, high pressure processing caused protein denaturation in egg white, which manifested in changes of its DSC-thermogram and intrinsic tryptophan fluorescence. Electronic nose testing showed changes of the head-space volatile composition of egg albumen, particularly as a function of radiation treatment. Both treatments caused changes in the NIR-spectrometric “fingerprint” of the liquid egg white. Various chemometric analyses of the results of the latter instrumental methods, particularly statistical techniques developed by the group of one of the co-authors of this article, demonstrated the potential for detection and characterization of the applied non-thermal processing techniques on liquid egg white. Irradiation induced more carotenoid degradation and lipid oxidation in liquid egg yolk than pressure processing.
Authors:I. Novák, É. Zámbori-Németh, H. Horváth, Zs. Seregély, and K. Kaffka
Oregano is used worldwide both as spice and crude drug, which is mainly provided by species of Origanum genus. The quality of the product is usually determined by chemical analysis, whereas in food industrial applications sensory tests are also practised. The aim of the present study was a comparison of parallel quality investigations of oregano samples by a new and effective instrumental sensory evaluation method, the “electronic nose”, and by gas-chromatographic and human sensory analysis. The GC analysis of essential oil components revealed mainly differences between plant species (Origanum vulgare subsp. hirtum and Origanum majorana). Main components of the oil of the former taxon are carvacrol and thymol, while those of marjoram are terpinene-4-ol, ?-terpinene and terpinolene. A wholesale oregano sample showing considerable divergence from the other ones with respect to ratios of carvacrol, ß-caryophyllene ß-cubebene and thymol. It was assumed not to belong to ssp. hirtum. The electronic nose analysis, evaluated by PCA, proved to be an appropriate, rapid, non-destructive, reagent-less method for the reliable separation of all of the oregano samples based on their complex aroma features. Assumptions could be made about correlations between separation of samples by the instrumental sensors and proportions of terpenoid compounds of the oil established by GC in some cases only. The varying essential oil content of the samples did not influence the success of instrumental evaluation. The instrumental and human sensory analysis showed similar results: varieties of O. majorana could be well distinguished on the basis of their complex aroma, while their gas-chromatograms did not show characteristic differences. The results call the attention that quality evaluation of drug items of aromatic plants should be oriented in different directions, considering the current utilisation area of the items.