34198

Fraud detection: Non-destructive determination of chicken fillet adulteration by MicroNIRTM Spectrometer

Rita Boerrigter-Eenling 1 ; Yannick Weesepoel 1 ; Eleftheria Tzoumeleka 1 ; Chang Hsiung 2 ; Chris Pederson 2 ; Nada O'Brien 2 ; Saskia van Ruth 3
Favorite this paper

Local production and import of chicken fillets in Europe and the Netherlands is a multimillion euro industry. Meat products may be susceptible to fraud, because economical gain can be substantial. As moisture and protein are the major value components of chicken fillets (on average a total of 98%), regulations concerning authenticity mainly focuses on the determination of these parameters. The EC regulation 543/2008 describes the official methods for the moisture and nitrogen (Kjeldahl, protein) content (NEN-ISO 1442:1997 and NEN-ISO 937:1992)1. The ratio of both parameters should not exceed a maximum value as has been set in the EC regulation. Both determinations are performed using homogenized chicken samples. A severe disadvantage of this procedure is that the total analysis procedure is lengthy (minimum 2 days) and it also restricts the number of chicken fillet samples that can be checked. A second issue, adulteration often remains undetected when looking solely to the moisture – protein ratio. For example “thawed fillets” can be offered as “fresh chicken fillets” in supermarkets or moisture and accompanying moisture retaining agents can be added, while not being declared on the package and therefore deceiving the consumer.
Near-infrared spectroscopy (NIRS) has been performed successfully before on homogenized meat samples for protein, moisture and fat determination as an alternative fast method2-4. However, this has rarely been performed non-destructively and with a functional portable device. Therefore in this study, the MicroNIRTM 1700 Spectrometer was explored as a fast tool for non-destructive determination of fresh-thawed fillets, presence of foreign additives (salt and porcine gelatine) and the moisture – protein ratio. 130 Samples were analysed and 5 scans per sample were taken. Due to the inhomogeneity of the samples, ‘classical’ principal component classification and regression methods show little correlation with the laboratory values. To overcome this problem, self-learning algorithms were exploited. In this case, Support Vector Machine classification and regression gave accurate results. For classification, false negatives (1%) and false positives (3%) were in an acceptable range after extensive external validation. Furthermore, for moisture-protein quantification, low deviations from the actual ISO values were found (2%). Current results show that on-site inspection of chicken fillets is a feasible application before proceeding to costly laboratory analysis of suspect samples.

References
1. EUR-Lex, European law. Last visited 14th October 2014. http://eur-lex.europa.eu/LexUriServ/LexUriServ.do?uri=OJ:L:2008:157:0046...
2. Anderson, S. Determination of Fat, Moisture, and Protein in Meat and Meat Products by Using the FOSS FoodScan™ Near-Infrared Spectrophotometer with FOSS Artificial Neural Network Calibration Model and Associated Database: Collaborative Study. Journal of AOAC International, 2007. 90 (4): p. 1073–1083.
3. McElhinney, J., Downey, G., and Fearn, T., Chemometric processing of visible and near infrared reflectance spectra for species identification in selected raw homogenised meats. Journal of Near Infrared Spectroscopy, 1999. 7(3): p. 145-154.
4. Cozzolino, D. and Murray, I., Identification of animal meat muscles by visible and near infrared reflectance spectroscopy. LWT - Food Science and Technology, 2004. 37(4): p. 447-452.