Application of NIR for monitoring quality traits of mechanically separated poultry meat

Increased demand for processed meat products as well as work towards reducing food waste have driven major technological improvements in processing equipment for harvesting of residual meat from poultry carcasses. The mechanically separated meat (MSM) obtained from these processes vary substantially in textural and biochemical characteristics, depending on carcass part used, separation technique and pressure applied during processing. Therefore, there is a need to develop an objective measure for classifying MSM quality which can be used by industry and food authorities. The EU project MACSYS which brings together nine research and industrial partners from five countries was initiated in January 2014 to develop such a measure. One objective of the project is to assess the potential of NIR for fast monitoring of fat, protein, moisture, calcium, and muscle cell integrity of MSM during production.
Nineteen MSM samples, representing a wide range of input carcass parts, separation techniques and pressures, two hand-deboned samples and samples of minced meat from breast fillet and drumstick were investigated. The samples were stored frozen and thawed at five degrees for 24 hours before measurement on a Bomen QFA flex 600 FT-NIR instrument (Q-interline A/S, Denmark) equipped with an InGAs detector. Each 100 g sample was measured 2 x 10 times in the range 11000-4000 cm-1 with a resolution of 16 cm-1 and 64 scans per measurement. The samples were turned in between each measurement to take into account the inhomogeneous nature of comminuted meat. Up to six replicates was measured per sample. Fat was measured by acid hydrolysis and solvent extraction, protein by Dumas combustion, moisture content by oven drying at 101 °C, and calcium by ICP-OES after acid digestion. Finally, muscle cell integrity was quantified by two different histochemical methods for which the results were combined into a reference measure.
PLS models were developed for prediction of calcium, moisture, fat and protein by NIR. Piece-wise multiplicative scatter correction (PMSC) with a window size of 46 and the 1. derivative were used as preprocessing for all models except protein, for which PMSC was used. The models were validated by cross-validation leaving out all measurements from one sample at a time. The resulting models had R2CV of 0.57, 0.75, 0.94 and 0.91 and RMSECV of 22.7, 0.99, 1.42 and 2.03 for calcium, protein, moisture and fat, respectively. The average coefficient of variation in the range 914-1064 nm for repeated measurements of the same sample proved to be the measure that best resembled the data trends found for muscle cell integrity as it could separate samples with a low degree of muscle degradation from the rest. However, more analysis is needed to fully understand if NIR can be used for predicting the muscle cell integrity. The ability to use NIR for estimating the calcium content seems difficult, while use of NIR as a fast method for prediction of fat, protein and moisture content of MSM seems straightforward. Validation of the results for fat, moisture and protein in a production setting, including more samples is in the planning.