Quality assessment of Italian dry-cured hams by FT-NIR spectroscopy

Dry-cured ham is one of the most representative typical meat products in Italy. It is a complex product, since processing technologies as well as raw material characteristics contribute to its quality. The modifications of characteristics such as pH and water activity, together with proteolytic and lipolytic reactions, produce changes that give rise to the typical characteristics of the final products. However, nowadays, determinations of cured ham quality parameters is time consuming and can require complex methods of analysis. Therefore, this paper investigates the suitability of FT-NIR combined with chemometrics for the assessment, in a single and rapid measure, of the most important quality features of Parma type dry cured-ham.
The research was carried out on 40 dry-cured hams after 17 months of ripening, obtained from just as many pig thighs (Italian Duroc × Italian Large White cross genotype). Before analyses, samples, obtained from a whole central slice of the ham, were finely ground to obtain a matrix as homogeneous as possible.
The samples were analysed by a FT-NIR spectrometer (MPA, Bruker Optics). Spectra were collected in the 12,500-4,000 cm-1 range (resolution of 16 cm-1 and 64 scans for sample and background) in diffusive reflection by an integrating sphere. Physico-chemical and chemical characteristics (water activity, pH, moisture, fat, protein, NaCl, TBArs, non protein nitrogen - NPN, total volatile nitrogen basis - TVB-N) were determined following official methods or commonly applied published procedures.
PLS regression models, between spectral data and physico-chemical and chemical variables, were calculated and cross-validated after pre-processing the spectra with Standard Normal Variate (SNV) combined with smoothing or first derivative (PLS_Toolbox, Eigenvector Research Inc., working under MatLab v. 7.4, The MathWorks).
Water activity, ranging from 0.863 to 0.937, was well predicted by PLS model on spectra pre-treated with SNV and first derivative achieving small errors (RMSEC 0.001, RMSECV 0.002). The best model achieved for pH prediction had RC2 and RCV2 of 0.964 and 0.885, respectively and RMSE lower than 0.003. Good regression models were obtained for moisture content being RC2 and RCV2 0.974 and 0.943, respectively. Acceptable models were obtained for NaCl (3.67-5.26 g/100g) prediction after spectral transformation by SNV and first derivative; even when RCV2 were not high (0.754), quite small RMSE were calculated. Fat content (12.54 -36.85 g/100g) and TBArs (ranging from 0.96 to 3.54 mgMDA /kg) models performed well reaching RMSECV lower than 0.18 % and 0.04 mgMDA /kg, respectively. Regression model for protein (ranging from 22.93 to 32.90 %) obtained after SNV and first derivative transformation was acceptable being RCV2 and RMSECV 0.75 and 0.18 %, respectively. The best performance was obtained for PLS models calculated for proteolytic indexes, NPN (19.98 – 33.17 %NPN/N) and TVB-N (21.02 – 33.17 mg/100g), as we obtained RCV2 of 0.936 and 0.975, respectively.
In conclusion, the present work demonstrated that FT-NIR performed as a good tool for predicting the main quality parameters of typical Italian dry-cured ham. FT-NIR guarantees fast and simple performance of analysis and substitutes several laborious analysis with one spectral acquisition.