ELECTRONIC NOSES APPLICATIONS AND TECHNOLOGIES
The electronic nose applied to food analysis
Capone, S.*; Distante, C.; Francioso, L.; Presicce, D.; Taurino, A.M.; Siciliano, P.; Zuppa, M.
Istituto per
la Microelettronica ed i Microsistemi, Sezione di Lecce, via Monteroni
Campus Universitario-Palazzina A3, 73100 Lecce, Italy*
Fax: +39 832 422552, E-mail: simona.capone@le.imm.cnr.it
Received November 05, 2004. In final form June 06, 2005
Abstract
There is currently a great interest in developing new techniques
for food quality assessment. Electronic nose is considered an attractive technique
for evaluating food aroma. In this work we present an overview of application
examples of an electronic nose based on sol-gel metal oxide gas sensor array
for the assessment of some foodstuffs (milk, olive oil and peach fruit). The
responses of the sensor array to the flavours of these foodstuffs have been
correlated with the results obtained by other standard techniques.
Resumen
Existe actualmente un gran interés en desarrollar nuevas
técnicas para la valoración de la calidad de alimentos. La nariz
electrónica se considera una técnica atractiva para evaluar
el aroma de alimentos. En este artículo presentamos una descripción
de los ejemplos del uso de una nariz electrónica basada en un arreglo
de sensores de gas del tipo óxido de metal (obtenido por el método
de sol-gel) aplicados a algunos comestibles (leche, aceite de oliva y duraznos).
La respuesta de la matriz de sensores a los aromas de los alimentos ha sido
correlacionada con los resultados obtenidos por otras técnicas estándares.
1. Introduction
Food quality can be defined as being “a combination of attributes or characteristic
data of a product, which have a significance in determining the degree of acceptability
of the product by the user”. At present there are great efforts in developing
new techniques for assessing food quality. Quality control of product fragrance
and flavor is based on the comparison of sensory, instrumental, analytical,
and if necessary microbiological data, with standards and specifications. Chromatographic
and spectroscopic analytical techniques (and their combination such as GC-MS)
and sensory analysis are commonly used. The “aroma” of a foodstuff
is a complex mixture consisting of thousands of different chemical volatile
species (chemical patterns). Analytical techniques allow chemical patterns
to be separated and the individual components identified and quantified. However,
these techniques are complex, expensive, time consuming and they require a well-equipped
analytical laboratory and a skilled staff. On the other hand, sensory analysis
provides the evaluation of foodstuff as a whole as deriving from the impact
of its odors/flavors on human senses. Again, it’s very costly to maintain
a skilled sensory panel, and there is a limit to the number of replicate samples
which can be evaluated due to olfactory adaptation to odors.
Some new trends are occurring in food quality assessment techniques, based on
the use of non-destructive spectroscopic techniques and electronic noses [1-5].
The latter are based on arrays of solid state gas sensors, mainly chemoresistive
metal oxide gas sensors (MOX) or quartz microbalance gas sensors (e.g. QMB).
The gas sensor array, exposed to the complex “chemical pattern”
of a foodstuff, analyze it as a whole, without decomposing it as occur for chromatographic
techniques, giving a simply pattern of sensor responses as read-output. Electronic
noses can find applications in food industry for quality control, process monitoring,
freshness evaluation, shelf-life investigation and authenticity assessment.
Considerable work has already been carried out on milk and dairy products [6-8],
coffee [9], oils [10-12], wine [13,14], tomato [15], meat and salami [16,17],
fruit [18-21], fish [22,23], and on many other foodstuffs and beverages, as
well as on the odor quality evaluation of food packaging [24].
Historically, the concept of an electronic nose arose from the need to
overcome the poor selectivity of chemoresistive gas sensors by linking a number
of different non-selective chemoresistive sensors in array configuration and
exploiting the unavoidable cross-sensitivity of the sensors. The term ‘array’
is not to intend to mean a particular spatial arrangement of the sensors in
the device, on the contrary, it refers to a vectorial organization of the output
of the single sensors under exposure to the same environment, so that the contribution
of each array element is a component of a multidimensional general problem.
The signals of such a sensor array can be processed by means of methods of Pattern
Recognition (PARC) which, exploiting the cross-correlations between the sensor
responses, extract information contained in the sensor-output ensemble [1].
In particular a PARC procedure performs a qualitative analysis of the environment;
it evaluates the multidimensional data-set from gas sensor array seeking the
underlying main relationships in the data-set itself just in order to analyze
the data structure and discriminate between different data classes (clusters)
belonging to different ‘chemical patterns’ (classification).
Moreover, a PARC procedure has to assign a sensor array output from an unknown
gas/odor to a class recognizing in such a way the occurrence of a particular
chemical pattern (identification). Of course, such multisensor systems
need calibration; by presenting many different chemicals to the sensor array,
a database of patterns is built up. This database of labeled patterns is used
to train the pattern recognition system. The goal of this training process is
to configure the recognition system in order to classify and eventually quantify
each chemical compounds in a gas mixture.
In this work a brief overview of the applications of an electronic nose based
on a metal oxide gas sensor array to the evaluation of some foodstuffs as milk,
olive oil and peach fruit is shown. The responses of the sensor array to the
flavors of these foodstuffs are correlated with the results obtained by other
standard techniques. In particular, we used microbiological and SHS/GC/MS analysis
for the assessment of dry salami, visible/near-infrared spectroscopy and destructive
techniques as refractometer and penetrometer for the evaluation of the total
soluble solid content (SSC) and firmness in peach fruit respectively, and finally
SHS/GC/MS analysis and pH measurements by standard pHmeter for the evaluation
of milk used for the production of mozzarella cheese. The aim of the work was
to demonstrate that the Electronic Nose can be a useful complementary tool,
not totally alternative, to conventional techniques in specific applications
of food industry. Indeed, this type of electronic device gives comparative rather
than quantitative information, hence, it is ideally suited for quick quality
control/quality analysis (QC/QA) checking.
Figure 1: Image of the
portable Electronic Nose developed
at IMM-CNR - Lecce (Italy).
2.1 Experimental
A portable
Electronic Nose has been used for the headspace analysis of the foodstuffs considered
in this work. The device has been entirely designed and realized at the IMM-CNR
- Lecce for application in food industry and environmental monitoring. It is
small, compact and completely automatic. The developed Electronic Nose instrument
(12x15x30 cm sized) consists of an electronic subsystem composed of an embedded
PC (5 stack modules in PC/104 format) and 2 electronic boards for the control
of sensors (an array of 8 metal oxide-based gas sensors and 1 digital temp/humidity
sensor) hosted in a stainless steel cell. The sensing elements are metal oxide
thin films prepared by the sol-gel technique, that is a wet chemical method
known to be suitable for the preparation of nanoscaled systems. In particular
the sensing materials are undoped and Ni, Os, Pt, Pd, Rh-doped SnO2
and undoped and Fe-doped In2O3 thin films. Each sensor
is deposited on an alumina substrate 2 x 2 mm2 sized and provided with a Pt-resistive
type heater on the backside and Au-interdigitated electrodes (50 µm spaced
fingers) on the front side. The sensors are bounded on rectangular standard
sockets and introduced into the test cell of the device. Internet and wireless
connections are also available. A flow of ambient air, previously dried and
cleaned by suitable filters, is extracted by a micropump and two electrovalves
designed to deviate the air flow into the sample vial for the collection and
the transfer of the sample volatile compounds into the sensor array cell. In
Figure 1 an image of the instrument is reported.
3. Results
and discussion
3.1 Milk
One of the most important technological tasks in the production of dairy
products is the quality control of starting milk. The chemical analysis of flavors
in dairy products is complicated by the heterogeneous nature of milk. In fact,
significant levels of lipids, proteins and carbohydrates in milk make difficult
to separate flavor-active chemicals based on general properties like polarity
or volatility. It is also very difficult to separate smell-active compounds
from the multiple phases found in milk, especially if the aim is to obtain an
extract, fraction or aliquot that is both enriched and representative. It is
reported in literature [25] that a lot of volatile organic compounds (VOCs)
are present in the headspace of milk (acetone, hexanal, 2-pentanone, 2-butanone,
toluene, limonene, heptanal, etc.) and their concentration depends on the nature
and thermal treatment of milk (UHT, pasteurized, sterilized, etc.). Furthermore,
a high quantity of water is also present, so the headspace turns to be represented
by a complex mixture of volatile organic compounds in an atmosphere containing
a high percentage of relative humidity. Analysis of milk volatile compounds
by electronic noses has been shown to be a promising tool to get information
on milk quality and to distinguish among different types of milks mainly in
relation to the thermal treatment [6, 26].
Here, since milk freshness is of primary importance for the milk processing
industry, we tested the aptitude of our EN to distinguish two different types
of milk (commercial UHT and pasteurized milk) as a function of their storage
days when its rancidity becomes relevant. For both types of milk, different
gas-sensing tests at different ageing days have been carried out by measuring
the sensor response to the headspace of 10 ml milk in 20 ml vial on a daily
basis. The data acquired by the EN-measurements were collected for a period
of 8 days for UHT and 3 days for pasteurized milk. Principal Component Analysis
(PCA) was then performed on the normalized data responses of the array for each
kind of milk. In both cases the clusters related to subsequent days of ageing
are well separated, so that it is possible to follow the evolution of rancidity
of milk during time. As an example, in Figure 2 the PCA
score plot for the pasteurized milk sample is reported. It can be easily observed
that the clusters evolve along preferential directions according to the different
degree of rancidity. The obtained results suggest an interesting application
of the electronic nose in dairy farming for the milk quality control measured
on milk ageing and rancidity.
In next stage, we evaluated the application of the Electronic Nose in the qualitative
analysis of the aroma of a mozzarella cheese produced in a local dairy (at Brindisi,
Italy) [27]. Since the quality of mozzarella depends on the composition and
types of basic milk as well as on the production process, we exposed the sensor
array to the volatile compounds of both, mozzarella samples and dairy parcels
of basic milk. In this dairy farming, milk parcels, coming from different stockbreeders,
are daily mixed in a single tank to form the basic milk (milk “massa”)
that is used for the production of mozzarella cheese. In particular, we analyzed
by EN stocks of milk of different stockbreeders –labelled as “Gallone,
Bruno, Argentiero, Gioia, Simeone and Massa”. PCA performed on sensor
array response gave satisfactory results concerning to the discrimination among
the different milk parcels (Figure 3). Moreover, the responses
of the sensor array to the volatile compounds of milk samples have been coupled
with the value of pH measured with a pHmeter. Multiple regression methods (MLR,
PCR, PLS, Surface Response Modeling) have been used to correlate the sensor
signals to the pH value that has to be predicted. Surface Response Modeling,
that is a polynomial regression method, applying the MLR method and a quadratic
ANOVA, gave the best prediction results since it is more suitable to account
for the non-linear behavior of sensors (Figure 4).
After the analysis by EN, a SHS\GC\MS analysis was carried out on the same milk samples kept at room temperature. This kind of approach allowed us to optimize an analytical method for the headspace analysis of milk samples by SHS\GC\MS which is comparable with the headspace sampling method used for the analysis with the electronic nose. In this way the array response to milk flavors can be correlated to the results of SHS\GC\MS analysis. As expected, the gas chromatograms of milk samples showed not many peaks due to the particular HS conditions (Tvial = 46 °C for 15 min); the choice of these conditions was imposed by the need to correlate the GC-MS results to those of Electronic Nose. However, since the gas-chromatogram pattern constituted a fingerprint of the headspace milk sample, the separation between the data clusters related to the different milk parcels in the PC-space can be justified by the composition in percentage of the milk sample headspace.
3.2
Olive oil
Olive oil has a fundamental importance in the ‘Mediterranean
diet’ and in the market, therefore there is a great interest towards all
the techniques able to provide an analysis of olive oil quality. Olive oil was
the first foodstuff to be classified by both chemical and sensorial analysis
according to the EU Normative. It means that an olive oil can be labelled as
extra-virgin only if both, chemical and sensorial characteristics, are within
certain standards established by law. Electronic Nose can be a promising tools
for olive oil industry that requires urgently reproducible, reliable, inexpensive,
easy to train and to use, objective ‘sniffing’ electronic device
dedicated to olive oil for different applications, the main being classification
and degradation studies. Authenticity is an issue of major concern across the
oil industry. Olive oil is marketed based on acidity grade in ‘olive-husk
oil, olive-oil, virgin olive oil, extra-virgin olive’. Olive oil is also
marketed based upon country and region of production and upon olive type. There
is a very large difference in price among, for example, Italian and Spanish
or Greek olive-oil.
However, it should be considered that the assessment of olive oil by an electronic
nose is not a simple task due to the large difference between the compositions
of the oil and its headspace. Indeed, those substances, on which fragrance and
quality are based, are poorly volatile, while other compounds (such as ethanol
and methanol) which are present at low concentration in the liquid phase, due
to their low boiling point, are present at high concentration in the headspace.
Figure 2: PCA score plot related to the sensor array response
to the volatiles of the pasteurized milk sample in different aging days.
Figure 3:
PCA score plot related to the sensor array response to the volatiles of different
stockbreeders milk samples – labeled as “Gallone, Bruno, Argentiero,
Gioia, Simeone and Massa” as provided to an Italian dairy industry.
Figure
4: Measured versus predicted pH values of milk samples
based on a quadratic regression method (Surface Response Modeling).
Figure
5: PCA score plot related to the sensor array response
to the volatiles of some commercial and non-commercial local olive oils from
Salento province (Italy).
Figure 6: PCA score plot related to some commercial olive
oils including an extra virgin olive oil derived from a biological cultivation.
The authors applied the developed Electronic Nose for testing a number of different commercial olive oils of various quality produced in different areas in Italy [10, 11]. Furthermore, the analyzed olive-oils were compared with local oils of the Salento region in Apulia ( Italy ). Indeed, one of the most important goals in the field of olive-oil characterization is their classification on the basis of their origin region. This is most important for the regions exhibiting the so called D.O.P. (Protect Origin Denomination) label which is an index of the quality of the product. A significant result was obtained analyzing the array response to the volatile compounds of some commercial and non-commercial local olive oils from Salento by PCA. In the score plot it can be clearly distinguished all the different classes of olive oils and the discrimination between the commercial and non-commercial local olive oils (Figure 5). In Figure 6 the PCA score plot, related to some commercial olive oils including an extra virgin olive oil derived from a Salento biological cultivation, is reported. Also in this case, the PCA representation gave an idea of the good ability of the sensor array to distinguish among olive oils of different quality, so that in the future pattern recognition techniques could be applied.
Figure 7: 3D-PCA score plot related to the sensor array response to the volatiles of different Tuscan single-cultivar EVOOs
Upon evaluation
of the quality of extra-virgin olive oils (EVOOs) it should be considered that
they are produced by mixing oils from different “cultivar”, that
is, different varieties of tree species coming from different geographical origin
areas of cultivation. The organoleptic properties of olive oils also depend
on the particular blend. In order to satisfy consumer requirements, oil from
a certain producer must be easily distinguishable and identified by presenting
the same smell as well as the same taste and color. On the other hand, for olives
it is of great importance to evaluate and conserve their existing genetic diversity,
still preserved in spite of the disturbance of the environments where they are
cultivated. The origin and the geographical distribution of such high variability
in the cultivated olive are still under investigation. Most studies agree that
after a spreading of new varieties of olive along the Mediterranean basins,
the majority of modern cultivars are derived from the crossing of these ancient
cultivars between themselves or by their breeding with plants followed by local
selection practices.
Hence, by considering this particular issue in olive oil industry, the Electronic
Nose was used to analyze different “single-cultivar” EVOOs produced
by Tuscan producers. In order to study a relation between the sensor responses
and the real head-space of olive oil samples, analytical technique like Headspace
Solid Phase Micro Extraction \ Gas Chromatography \ Mass Spectrometry (HS-SPME\GC\MS)
analysis was applied to the analysis of volatiles compounds in EVOOs samples.
The obtained GC-MS data were used to identify the particular compounds and characterize
the chemical composition of the EVOOs samples. Statistical analyses were carried
out on data obtained from Electronic Nose and GC-MS.
Figure 8: Histogram showing the different chemical compounds identified by HS-SPME\GC\MS analysis in EVOOs aroma.
In particular, twelve samples of different Tuscan single-cultivar EVOOs (labeled as Leccino, Maremmano, Olivastra di Montalcino, Gremigno di Montecatini, Leccione, Madonna dell’Impruneta, Lazzero di Prata, Americano, Scarlinese, Morcone, 2CRL, Tisignana) were analyzed by means of EN and HS-SPME\GC\MS [28]. At first, PCA analysis based on the sensor array data, gave a not full satisfactory discrimination among the different EVOO samples. Figure 7 reports the 3-dimensional score plot in which the drawn ellipses did not allow to identify clearly a complete separation among the clusters. By means of HS-SPME\GC\MS analysis, only eighteen different chemical compounds were identified in the EVOOs aroma. The histogram in Figure 8 summarizes these results. From a qualitative point of view, from HS-SPME\GC\MS analysis one can see that aldehydes, terpenes and alcohols are the main chemicals families present in the analyzed headspaces. Aldehydes, in particular hexanal and nonanal, coming from lipid oxidation, are the predominant chemical family present in the headspace.
3.3 Peach
Evaluating fruit ripeness has long been a fundamental issue in postharvest technology
since ripeness is perceived by customers as the main quality indicator. Standard
physical-chemical techniques enable to measure some quality indicators, as total
soluble sugars content (SSC) and firmness, but they require the destruction
of the fruit samples used for the analysis. This is why, nowadays, predictions
of shelf-life ripeness state is mainly based on sensory analysis and practical
experience, although they may be influenced by subjective evaluations.
Figure
9: Schematic diagram of the experimental setup for NIR testing of peach
fruit.
Figure 10: Best prediction results for EN measurement (firmness assessment) and for NIR measurement (SSC assessment).
Aroma analysis
by Electronic Nose is to be considered an important index of fruit ripeness
assessment. The formation of aroma compounds in fruits is a dynamical process,
because volatile compounds are continuously synthesized and developed during
fruit growth and maturation. Moreover, Near-Infrared Spectroscopy has been shown
to provide rapid and non-destructive measurements of internal properties as
SSC due the absorption bands of carbohydrates in this wavelength region.
We evaluated the performance of an Electronic Nose and Near-Infrared Spectroscopy
as additional, non-destructive and objective techniques in the analysis of ripeness
of post-harvested peaches [19, 20]. With this aim the electronic nose responses
to the aroma of intact peach samples and the related NIR spectra have been coupled
with the values of total Soluble Sugars Content (SSC) and firmness measured
by a refractometer and a penetrometer respectively. Multivariate regression
methods (MLR, PCR, PLS, Surface Response Modeling) have been used to correlate
the sensor signals and the NIR spectra to the fruit quality indicator to be
predicted.
A set of several peaches at different maturity degree was analyzed by Electronic
Nose and NIR spectroscopy. Figure 9 shows the schematic
diagram of the setup for NIR testing of peach fruit. After the analysis by the
Electronic Nose and the NIR spectrometer the peach samples were destroyed to
measure the SSC and the firmness. Total Soluble Sugars Content (SSC) is a major
characteristic used for assessing peach fruit quality. Its measurement is normally
performed destructively on juice. Two slides of flesh were cut from two opposite
sides of the sample and the mixed juices were squeezed directly onto a refractometer
which gives the medium value of SCC in °Brix. Firmness is one of the easiest,
fastest and cheapest methods to assess ripeness and also one of the better correlated
with ripeness parameters. It was measured with a penetrometer (8 mm tip) on
the two opposite wider sides, at the center of each fruit. The instrument measures
the force required for the mechanical probe to penetrate into the tissue of
unpeeled fruit. The measured firmness values were averaged and expressed in
Kgweight.
By means of multivariate regression techniques, relationships were established
between Electronic Nose response to aroma peach and fruit quality parameters
(SSC and firmness). Firmness was more precisely predicted than SSC. The regression
technique based on a quadratic regression (i.e. Surface Response Analysis) gave
better prediction results respect to the linear regression techniques. Excellent
prediction of the quality indicators were also obtained by NIR spectroscopy
by simple linear regression techniques (i.e. PLS). As example, Figure
10 showed the best results for EN measurement (firmness assessment) and
for NIR measurement (SSC assessment).
4. Conclusions
In this overview of possible applications of Electronic Nose in food industry,
the responses of the sensor array to the flavors of some foodstuffs with the
results obtained by other standard techniques were compared. It was demonstrated
that the Electronic Nose, when properly used, can be a comparative and complementary,
not totally alternative device to conventional techniques in specific applications
for food industry. In the future, the authors´ aim will be to join together
the information acquired by the electronic nose with other techniques (as Vis/NIR)
data to take advantage of the synergic effect expected by the use of different
techniques.
References
[1] Gardner, J.W.; Bartlett, P.N., Electronic Noses–Principles and Applications, Oxford University Press, ISBN 0 19 855955 0, 1999.
[2] Mielle, P., Trends in Food Science and Technology, 1996, 7, 432.
[3] Schaller, E.; Bosset, J.O.; Escher, F., Lebensm.-Wiss. u.-Technol., 1998, 31, 305.
[4] Göpel, W., Sensors and Actuators B, 1998, 52, 125.
[5] Stetter, J.R.; Strathmann, S.; McEntegrat, C.; Decastro, M; Penrose, W.R., Sensors and Actuators B, 2000, 69, 410.
[6] Ampuero, S.; Bosset, J.O., Sensors and Actuators B, 2003, 94, 1.
[7] Brudzewski, K.; Osowski, S.; Markiewicz, T., Sensors and Actuators B, 2004, 98, 291.
[8] Capone, S.; Epifani, M.; Quaranta, F.; Siciliano, P.; Taurino, A.; Vasanelli, L., Sensors and Actuators B, 2001, 78, 174.
[9] Costa Freitas, A.M, Parreira, C.; Vilas-Boas, L., Journal of Food Composition and Analysis, 2001, 14, 513.
[10] Quaranta, F.; Rella, R.; Siciliano, P.; Capone, S.; Distante, C.; Epifani, M.; Taurino, A., Sensors and Actuators B, 2002, 84, 55.
[11] Taurino, A.; Capone, S.; Distante, C.; Epifani, M.; Rella, R.; Siciliano, P., Thin Solid Films, 2002, 418, 59.
[12] Cerrato Oliveros, M.C.; Pérez Pavón, J.L.; Garcia Pinto, C.; Fernández, M.E.; Laespada, B., Moreno Cordero, B.; Forina, M., Analytica Chimica Acta, 2002, 459, 219.
[13] Buratti, S.; Benedetti, S.; Scampicchio, M.; Pangerod, E.C., Analytica Chimica Acta, 2004, 525, 133.
[14] Santos, J.P.; Arroyo, T.; Aleixandre, M.; Lozano, J.; Sayago, I.; Garcia, M.; Fernandez, M.J.; Ares, L.; Gutiérrez, J.; Cabellos, J.M.; Gil, M.; Horrillo, M.C., Sensors and Actuators B, 2004, 102, 299.
[15] Berna, A.Z.; Lammertyn, J.; Saevels, S.; Di Natale, C.; Nicolaï, B.M., Sensors and Actuators B, 2004, 97, 324.
[16] O’Sullivan, M.G.; Byrne, D.V.; Jensen, M.T.; Andersen, H.T.; Vestergaard, J., Meat Science, 2003, 65, 1125.
[17] Taurino, A.M.; Dello Monaco, D.; Capone, S.; Epifani, M.; Rella, R.; Siciliano, P.; Ferrara, L.; Maglione, G.; Basso, A.; Balzarano, D., Sensors and Actuators B, 2003, 95, 123.
[18] Brezmes, J.; Llobet, E.; Vilanova, X.; Saiz, G.; Correig, X., Sensors and Actuators B, 2000, 69, 223.
[19] Di Natale, C.; Macagnano, A.; Martinelli, E.; Proietti, E.; Paolesse, R.; Castellari, L.; Campani, S.; D’Amico, A., Sensors and Actuators B, 2001, 77, 561.
[20] Capone, S.; Distante, C.; Epifani, M.; Rella, R.; Siciliano, P.; A. Taurino, A., Analysis of peaches ripeness by an electronic nose and near-infrared spectroscopy, Proceedings of the 7th Italian Conference on Sensors and Microsystems, 2002, Bologna, Italy, in ‘Sensors and Microsystems’, copyright © 2002 by World Scientific Publishing Co. Pte. Ltd., p.125-130, ISBN: 981-238-181-3.
[21] Saevels, S; Lammertyn, J.; Berna, A.Z.; Veraverbeke, E.A.; Di Natale, C.; Nicolaï, B.M., Postharvest Biology and Technology, 2004, 31, 9.
[22] Olafsdottir, G.; Nesvadba, P.; Di Natale, C.; Careche, M.; Oehlenschläger, J.; Tryggvadóttir, S.V.; Schubring, R.; Kroeger, M.; Heia, K.; Esaiassen, M.; Macagnano, A.; Jørgensen, B.M., Trends in Food Science & Technology, 2004, 15, 86.
[23] O’Connell, M.; Valdora, G.; Peltzer, G.; Martín Negri, R., Sensors and Actuators B, 2001, 80, 149.
[24] Frank, M.; Ulmer, H.; Ruiz, J.; Visani, P.; Weimar, U., Analytica Chimica Acta, 2001, 431, 11.
[25] Mariaca, R.; Bosset, J.O., Lait, 1997, 77, 13.
[26] Sberveglieri, G.; Comini, E.; Faglia, G.; Neiderjaufer, G., Benussi, G.; Contarini, G.; Povolo, M., Seminars in Food Analysis, 1998, 3, 67.
[27] Dello Monaco, D.; Taurino, A.M.; Capone, S.; Distante, C.; Siciliano, P.; Rella, R., Qualitative analysis of local dairy products by an Electronic Nose and GC-MS, Proceedings of the 9th International Symposium on Olfaction and Electronic Noses (ISOEN 2002), Rome, September 29th - October 2nd, 2002.
[28] Dello Monaco, D.; Distante, C.; Presicce, D.S.; Siciliano, P.; Taurino, A.M.; Zuppa, M.; Cimato, A.; Sani, G., Sensors and Actuators B, 2005, In press, corrected proof, available online 10 August 2005.