An On-Line Near-Infrared (NIR) Transmission Method for Determining Depth Profiles of Fatty Acid Composition and Iodine Value in Porcine Adipose Fat Tissue

Abstract

The present work describes a measurement method using spatially resolved near-infrared (NIR) spectroscopy to determine porcine carcass fat quality as a function of the distance to the skin by estimating its iodine value (IV). The new method is capable of performing on-line carcass grading at full production speed (approximately 1000 carcasses per hour). The method is demonstrated in an experiment where 35 carcasses were sampled at an abattoir, selected from three feeding groups. The NIR transmission instrument was applied on the loin of each carcass, and a parallel reference sample was removed and processed into 1.8 mm thick disks, representing a depth-of-fat profile from the loin. The disks were analyzed for fatty acid composition using gas chromatography (GC) and for IV. A principal component analysis (PCA) of the obtained GC reference values clearly showed that the feeding regimes can be differentiated. Using interval partial least squares (iPLS) regression, a model was produced that can predict the IV of the fat at a given measured depth with a root mean square error of cross-validation (RMSECV) of 1.44. The results show how the IV varies as a function of feeding regime and as a function of fat depth. The maximum variation found within a single depth profile was 10.1 IV from the skin to the innermost part of the fat layers. In the sample material investigated the average span in IV between the average values of the two porcine backfat layers was 6.4 IV (the maximum difference was 8.6 IV). The new method can provide the abattoir with new chemical information about fat quality and production quality that will open new possibilities of meat/carcass grading and product development.

Iodine Value and Fatty Acids Determination on Pig Fat Samples by FT-NIR Spectroscopy: Benefits of Variable Selection in the Perspective of Industrial Applications

“…Hence, this sort of representation shows the degree of convergence of the various models computed for each variable. It can be observed that, for all the modeled chemical variables, the most frequently selected region is the one corresponding to the spectral bands centered at 5785 and 5670 cm −1 , in which the absorption bands related to the C-H stretching first overtone of CH 2 and CH 3 groups are located (Sørensen et al 2012;González-Martin et al 2005;Pérez-Juan et al 2010;Li et al 1999).…”

“…Finally, in the work by Li et al (1999) the bands near 8475 and 4675 cm −1 are attributed to strong -CH=CH-(cis double bond) overtone and combination band vibrations, respectively. Interestingly, most of the bands selected for calibrating IV were the same selected by iPLS in the work by Sørensen et al (2012).…”

“…In a recent paper concerning animal fat blends destined for biodiesel production (Adewale et al 2014), the IV values used as the response variable in the calibration models were measured using the Wijs analytical method. In other works (Prieto et al 2014;Gjerlaug-Enger et al 2011;Sørensen et al 2012), the IV reference values were calculated by an equation that takes into account the composition in fatty acids expressed in percentages (AOCS 1998;Pétursson 2002).…”

“…Conversely, in other works the fat samples were analyzed without any preparative step, using different methods: diffuse reflectance (Ripoche and Guillard 2001), transmission (Sørensen et al 2012) and fiber optic probes (Zamora-Rojas et al 2013;González-Martin et al 2005;Pérez-Juan et al 2010). Müller and Scheeder (2008) also measured the samples in diffuse reflectance but they homogenized the fat samples before spectra acquisition.…”

“…Concerning the prediction of iodine value and fatty acids based on NIR spectra, only in Sørensen et al (2012) the interval-PLS (iPLS) variable selection was applied in addition to the classical partial least squares (PLS) calibration method. However, variable selection should be often considered when dealing with FT-NIR datasets to identify and select the spectral regions in which the information of interest is located (Ferrari et al 2011;Foca et al 2009;Ulrici et al 2008;Cocchi et al 2005Cocchi et al , 2006.…”