Electronic Noses: Applications in the USA
By Rebecca N. Bleibaum, M.S. and Herbert Stone, Ph.D.
Summary
Using instruments to replace the human senses have been a research topic for more than a century, with the electronic nose as the current focus. In more recent times, Schutz, et al. (1961) and Wilkens and Hartman (1964) demonstrated the feasibility of an electronically based odor sensing system, early versions of electronic noses. Almost two decades past before contemporary electronic noses were described in the literature (Persaud and Dodd, 1982, Bartlett, et al., 1997). Advances in the development of sensor array technologies,
use of pattern recognition techniques, and neural networks, have had a significant impact
on the sensitivity and the utility of these instruments. Today they are capable of identifying, classifying, or assigning value judgments to the composition of a variety of products. Not surprisingly, there has been a growing interest in their applications in the food and beverage industry, in part because it represents a means of reducing reliance on the human judgment.
Both hand held and larger machines are being developed. However, there are a number of problems that remain to be addressed before the instruments will achieve wide spread use
in the food industry (Bartlett, et al., 1997). Problems that need more consideration include: sensitivity to specific chemicals, effects of moisture, extent to which an artificial system
can simulate human response behavior, and the importance of aroma in predicting consumer acceptance in relation to other sensory components such as taste and texture. None the less, interest in the use of these instruments remains high. This discussion considers the use of
the electronic nose in relation to contemporary sensory practices, and describes results of
a recently completed survey among users of electronic noses.
Background
The electronic nose (EN) represents an important advance in the use of odor analysis for
the evaluation of food and beverage quality. The development of instruments to measure food and beverage characteristics such as aroma, taste, and texture has a very long history. For example, measurement of texture using mechanical devices have been described in
the literature for more than 100 years. Bourne (1982) provided an excellent review of the subject describing mechanical devices to measure product changes, particularly the fruit ripening process, or to simulate the chewing process to measure meat tenderness. While
such devices have undergone substantive change in terms of sensitivity, measurement, design, etc., the basic principles have remained unchanged. Using a sensing system,
whether chemical and/or physical, enabled one to detect and record changes in a material
(in this instance a food or a beverage). Such a system increased understanding of the
human perceptual process, as well as reducing reliance on the human judgment, which
was described as variable and subjective. These systems enabled actions to be taken that would enhance finished product quality, and/or minimize presence of contamination.
The history of odor detection has always included both human and instrument systems.
The earliest instruments began with measuring total volatiles. This was followed by development chromatographic techniques and gas-liquid chromatography which enabled separation of chemicals as well as ability to separate specific chemicals at very low concentrations; e.g., ppm (parts per million) and less, to facilitate their identification. Subsequent developments enabled integration of the chromatographic separation with
even more sophisticated instrumentation such as mass spectroscopy to more precisely
clarify chemical structure, and so forth. During this same time period, Schutz, et al.,
(1961) and Wilkens and Hartman (1964) described their efforts to devise other systems
for the detection of odorants. Schutz and coworkers investigated use of specific enzyme systems associated with olfactory tissue that were reactive to odorants and correlated results with the physical and chemical properties of those odorants. At the time, the constraints
of separation technology, enzyme stability, and related issues impeded further development
of this approach. Wilkens and Hartman, on the other hand, took a very different approach
to the problem; they devised an instrument, an electronic analog that relied on micro electrodes made from combinations of metallic wires for the detection of odorants. As shown in Figure 1, this instrument was based on the concept of a polarizing process for
the micro electrodes, to reflect the olfactory receptor process. These authors were able to demonstrate differential response sensitivity to various volatile chemicals with just a few micro electrode metal combinations. This early research was "lost" as emphasis and
research support shifted to even more sensitive instruments.
Current Situation
Today's ENs represent significant advances in electrode specificity, sensitivity to
individual chemicals, use of computer chip technology for analyses, and in accessing
pattern recognition and neural network statistics as an aid in data interpretation (see Bartlett, et al., 1997, Lewis, 1996; Taubes, 1996). There is no question that current technology and combined with advances in understanding of the olfactory process, make the EN a very powerful tool. However, it has its limitations, as do most instruments of this type, especially if they are to be used in a manufacturing environment. As noted by Giese (1993) and Wilhelmsen, et al. (1998), on-line sensors represent a major business opportunity in the
food industry provided they can withstand the rigors of a factory, will not contaminate
the product being measured, and can take into account the inherent variability of the raw materials used in the manufacture of foods and beverages. In their review of ENs, Bartlett and coworkers noted that success will come about most likely through single system applications rather than broadly based applications. Considering the information developed in our survey, this assessment of applications is reasonable, provided current limitations can be overcome.
As previously noted, use of sensing systems represents an effort to replace and/or minimize the need for the human judgment in the evaluation of foods and beverages; whether for
food safety reasons, or in situations where response time is limited; e.g., during processing. Such applications are reasonable provided the aforementioned limitations have been addressed. However, efforts to directly measure food quality have a much greater challenge.
The Sensory Challenge
The literature on electronic noses emphasizes the importance of emulating the human olfactory system but with minimal or no consideration of the rest of the sensory system.
On the one hand, this focus on the olfactory system is understandable and necessary; it is only part of the picture in so far as concerns the measurement of food and beverage quality. The perceptual process is an integrated and interactive system; stimulation of one sense
can and will influence response to the other senses. For example, appearance can influence aroma, which in turn influences taste, and so forth (Stone and Sidel, 1993). An example of this is shown in Fig. 2, for three beverages, which depicts the important sensory attributes derived from a trained panel using the QDA™ method (Stone and Sidel, 1998). Determination of importance was derived from a regression analysis with the consumer preference judgment as the dependent variable. It can be seen that there were 13 attributes that were critical to product acceptance, of which sweet aroma was only one. For some products the number of important attributes may be fewer, but rarely is it a single attribute. In addition, the presence of a specific attribute is not reflected in an all or none situation. For example, presence of compound A at concentration X may result in product acceptance but at 10-X and at 10+X, product rejection. This should not be construed to mean that the electronic nose has no future, rather its commercial potential rests with those situations where there is clear evidence of direct or indirect marketplace relevance.
The Survey of Users
In August of 1998 we conducted a telephone survey with approximately 30 researchers
who expressed an interest in and use of electronic noses in the food and beverage industry. A copy of the survey and the results are shown in Tables 1 and 2. Of those surveyed, 41% were extremely familiar with their usage, 48% were very familiar; and only 11% were only somewhat familiar. Most researchers (77%) had more than 2 years of experience with the instrument, and 38% had 4 or more years. Most companies or departments were using
them in Research & Development (89%), while fewer were using them in the Chemistry
and Sensory Evaluation Departments (30% each). Quality Assurance (22%) and Quality Control (18%) groups also were familiar with them and were the only departments using them on-line (4%). The types of products currently being measured by electronic noses
were primarily raw ingredients (44%), followed closely by finished products (37%), with fewer using them for on-line intermediate production measures (22%). This was not an unexpected finding.
Respondents mentioned numerous types of products currently being measured with electronic noses. As reported in the literature (Lewis, 1996), it supported a wide range of applications that companies were considering. In addition to beverages (wines, fruit juices, waters), all types of snacks, seafood's, fruits and vegetables were being considered, along with such diverse markets as vehicle exhaust, breath alcohol levels, perfumes, chemical weapons, bombs, etc.
By far, the most common type of electronic nose currently in use was the electronic
polymer based (56%). Others of interest include metal oxide (8%), semiconductor (4%), and tin oxide (3%). For more information about these and other types of sensors, see
Freund and Lewis (1995) and Lonergan, et al. (1996).
Data generated from the electronic noses was primarily analyzed by statistical programs,
but only 30% were using artificial neural networks. Neural networks have been used to develop patterns or fingerprints of chemical vapors. In combination with sensor arrays, generally, the number of detectable chemicals is larger than the number of sensors (e.g., wine aromas). Of those surveyed, most systems required only 2 to 3 weeks to train operators; however 4% said it took longer than 4 months to train, indicating a higher level of complexity, primarily those using neural network systems.
When starting the investigation for electronic noses, few had 'very' or 'extremely high' expectations (31%), while the same % had low expectations. After using electronic noses, 36% were 'very' or' extremely' satisfied, while 28% were 'not very' or 'not at all' satisfied.
Shelf stable foods made up the majority of products evaluated by electronic noses in the food and beverage industry (37%), whereas fewer were using them for refrigerated (19%) and frozen foods (11%), and beverages (7%).
The relationship with consumer acceptance and trained panel tests have been studied by various companies, and 42% reported it was 'very' or 'extremely' related to consumer
and trained panel behavior, whereas 34% reported it was 'somewhat' or 'not at all' related.
Based on this survey, it was clear that there is considerable interest in usage of electronic noses in the food and beverage industry. Many companies have developed quite sophisticated research methodology to better understand the relationship to the information gathered from an electronic nose with both consumer acceptance and sensory evaluation panels. This interest and research will continue and it is likely that important applications will be identified.
Conclusion
Interest in the electronic nose is currently very strong as food and beverage companies investigate its potential for their products. It represents an opportunity to gain further
insight into food and beverage quality, to identify those volatile chemicals that are
directly or indirectly linked with product change. Such product change could be desired
or in other instances, undesired. In either situation, the use of such an instrument has the
potential to minimize product quality deterioration, thus increasing profitability.
However, such opportunities can only be accomplished when there is a good understanding of the relationship between specific chemicals and consumer behavior. At present such relationships are the exception, and considerably more needs to be accomplished to make
the electronic nose a practical system.
