Talanta 80 (2010) 1899–1906 Contents lists available at ScienceDirect Talanta journal homepage: www.elsevier.com/locate/talanta Threshold detection of aromatic compounds in wine with an electronic nose and a human sensory panel José Pedro Santos a,∗ , Jesús Lozano b , Manuel Aleixandre a , Teresa Arroyo c , Juan Mariano Cabellos c , Mar Gil c , Maria del Carmen Horrillo a a Instituto de Física Aplicada (CSIC), Madrid, Spain b Grupo de Investigación en Sistemas Sensoriales, Universidad de Extremadura, Badajoz, Spain c Department Agroalimentación, Instituto Madrile˜ no de Investigación y Desarrollo Rural, Agrario y Alimentario, Madrid, Spain article info Article history: Received 1 July 2009 Received in revised form 9 October 2009 Accepted 18 October 2009 Available online 28 October 2009 Keywords: Electronic nose Sensory panel Wine compounds thresholds abstract An electronic nose (e-nose) based on thin film semiconductor sensors has been developed in order to compare the performance in threshold detection and concentration quantification with a trained human sensory panel in order to demonstrate the use of an e-nose to assess the enologists in an early detection of some chemical compounds in order to prevent wine defects. The panel had 25 members and was trained to detect concentration thresholds of some compounds of interest present in wine. Typical red wine compounds such as whiskeylactone and white wine compounds such as 3-methyl butanol were measured at different concentrations starting from the detection threshold found in literature (in the nanograms to milligrams per liter range). Pattern recognition methods (principal component analysis (PCA) and neural networks) were used to process the data. The results showed that the performance of the e-nose for threshold detection was much better than the human panel. The compounds were detected by the e-nose at concentrations up to 10 times lower than the panel. Moreover the e-nose was able to identify correctly each concentration level therefore quantitative applications are devised for this system. © 2009 Elsevier B.V. All rights reserved. 1. Introduction Electronic noses have aroused in last years as a powerful tool in many applications mainly in food [1–3]. Their performances are usually compared with respect to analytical methods as gas chromatography–mass spectrometry or human sensory panels. Comparison with human panels has been performed mostly in food quality applications [4–6] with a qualitative approach. There are few examples of quantitative applications [7] and are outside this field. Electronic noses have been applied to wine related measure- ments since last decade for several purposes: varietal origin [8], oak toasting levels [9], regional characterization [10–12], ageing [13], defects detection [14–16] and sensory properties prediction [17,18] among others. A sensory panel can provide information about the sensory properties of consumer products that is far richer than can be provided by instrumental devices alone, but significant training is required before the panel becomes a reliable sensory instrument [19] specially dealing with wine, a very complex product. ∗ Corresponding author. E-mail address: josepe@ifa.cetef.csic.es (J.P. Santos). In the present work a sensory panel and an electronic nose (e- nose) based on semiconductor oxide sensors have been trained in parallel with several aromatic compounds present in wine in order to compare their performances in threshold detection and concentration quantification. This threshold is the minimum con- centration of an aroma in water perceived for at least 50% of the members of a sensory panel. Other term related to the olfactory threshold detection is the difference threshold detection, the mini- mum amount of an aroma that has to be added to a product already containing this aroma in order to produce an appreciable sensory change. This concept is especially useful in dealing with compounds in wine. Usually this threshold is evaluated in an artificial wine [20] but in the present case we used real wines. The great influence of water and ethanol in the response of semi- conductor oxide sensors is known [21,22]. We have employed two sampling methods, headspace and purge and trap in order to eluci- date the influence of these interferents in the e-nose performance. The aim of this work is to assess if the sensory panel can be substituted by an electronic nose for some specific tasks such as threshold detection. This experimental method may eventually replace current sensory evaluation techniques, with a simpler qual- itative and quantitative technique, offering better reliability and reproducibility of results, faster operation and lower cost. 0039-9140/$ – see front matter © 2009 Elsevier B.V. All rights reserved. doi:10.1016/j.talanta.2009.10.041