Raman Spectroscopy of Meat


Quick Links
Link to J Renwick Beattie's homepage
link to J Renwick Beattie's about me page
Renwick Beattie's page on consultancy and training for Raman spectroscopy, multivariate analysis and signal processing
link to J Renwick Beattie's publication list
link to J Renwick Beattie's resources page
link to J Renwick Beattie's links page
Link to J Renwick Beattie's Sitemap
contact J R Beattie

More information on the structure of meat and its proteins can be found in my introduction to the topic.

Raman spectra of muscle based food meat fish salmon lamb beef pork chicken hake whiting codFigure 1 shows an average Raman spectrum of a number of muscle based foods. The majority are dominated by the Raman spectrum of protein, with the exception of salmon which is dominated by polyunsaturated oil. Lamb, beef then chicken had progressively smaller amounts of more saturated fats, while pork and the white fish had no readily identifiable fat bands in their spectra. The main peaks of interest with regards to the proteins indicated on the spectrum of pork (which is predominantly protein), with the numbering here cross-referencing the peak assignments listed in Table 1, while spectra and assignments for fats and oils can be Raman spectroscopy pork peak assignments spectrum proteinfound in my introduction to fats.









Raman spectrum pork protein assignment












shear force variation with day of aging for pork

Raman spectroscopy pork aging tenderness PLSDA discriminant analysisThe pork samples were aged for between 3 and 14 days, being split into two groups and handled differently to induce a difference in the aging process that would give a wider range of toughness. Figure 3 shows the variation in the toughness of these pork samples, as measured by the shear force value. This shear force measures the amount of pressure required for a standardised blade to cut through a standardised piece of meat, i.e. how tough it would be to cut or chew. The CD group was prepared under commercial conditions, while the AB group were prepared in an experimental laboratory and it is clear that the experimental set is much tougher overall than the commercially prepared set. Also, there is little tenderisation in the AB group, but the CD group Raman spectroscopy pork PLS regression tenderness toughness shear force valueshows a significant drop in toughness after 11 days. Figure 4 shows the score plot generated from discriminant analysis (PLSDA) of the data. What is interesting is how closely the shape of the data reflects the variation in shear force shown in Figure 3. Although PLSDA was only given information about the day of aging (x axis in Figure 3) it has managed to replicate the toughness trend. The shear force measurement is unable to distinguish between samples that are tender due to incomplete contraction on day 3 and due to ageing on day 11, but there is a very clear difference between the two in the Raman data. Furthermore the difference in toughness due to the handling conditions is also readily distinguished by the Raman (which is why the CD set is shown separately here). Indeed, the Raman signal is strongly correlated with the shear force value of the pork, with an R2 of 0.77, with the scatter plot for the regression in Figure 5.







TheRaman spectra of pork spectroscopy cooking raw to cooked correlation discussed above was carried out between Raman signals recorded on raw meat and shear force measurements carried out on cooked meat. The cooking process has significant effects on the properties of meat, turning translucent pink pork chops to an opaque beige colour, inducing dramatic changes in texture and expelling juices from the meat. When the spectra from cooked meat was employed a similarly good correlation was obtained, which would be expected under the highly controlled cooking conditions employed in a laboratory situation. However, the cooking process is more divergent in the real world and the extent of cooking is a critical health issue. The Raman spectrum undergoes dramatic changes upon cooking, as shown by Figure 6. The biggest changes can be attributed to a transition from alpha helix to a beta sheet protein secondary structure upon cooking (blRaman spectroscopy of pork, PLS regression against cooking timeue vs magenta arrows). In addition there is a large number of changes in specific amino acid residue bands such as tyrosine, tryptophan and histidine. The scissor band at 1450 cm-1 is sensitive to the polarity of the protein environment, and this decreases upon cooking. The liquid expelled during the cooking process (the cooking loss) eliminates a lot of polar water and solutes so it is not surprising that this band is altered. Indeed the Raman signal can be correlated with the cooking loss (R2 = 0.71 for Raman spectra from raw, 0.72 from cooked samples). Figure 7 shows that the changes observed in the Raman signal are strongly correlated with the extent of cooking the sample has undergone, suggesting that Raman spectroscopy could prove a useful tool for determining this critical parameter.

However, the above results are all based on laboratory measurements and food is intended to be eaten by consumers. It is important that any instrumental method of quality assessment can ultimately be tied back to what the consumer actually wants. Consumer preference attributes are typically affected by a complex array of underlying biochemistry, biophysics and biomechanics. Shear force is unable to reliably reproduce all the subtleties of chewing, the physical properties of teeth, nor the mechanical movements of the jaws. Indeed people do not replicate each other in these processes either and even one person will not be completely consistent. As a result a taste test carried out by consumers is essential for determining the Raman spectroscopy PLS regression beef sensory tenderness overall acceptabilityefficacy of any quality assessment. Due to the variation that exists with a person and between people even taste tests are not great predictors of a second taste test, with inter-test correlation typically around R2 of 0.75. I have seen instrumental assessment techniques claim R2 over 0.9 against taste tests despite the fact this is theoretically impossible and therefore the results suggests serious over-fitting. 

beef sensory perception tenderness flavour aroma I carried out a correlation between Raman spectra and a taste panel assessment of beef. The panel assessed a number of key food quality characteristics in addition to rating the overall acceptability of the cut. In order to assess the relative contribution of each characteristic to the overall acceptability I regressed this against the other characteristics measured in the panel. Figure 8 shows the relative contributions of each characteristic to the overall impression of the beef. The study replicated the previously reported result that in the western world the most important factor was the acceptability of the texture of the meat. The acceptability of texture is contributed to by the perceived tenderness and juiciness which where both well correlated. Figure 9 shows the result of regressing the Raman spectra against each of the acceptability attributes. Raman spectroscopy was unable to predict the consumer preference for aroma or flavour, but was able to predict acceptability of texture and overall acceptability of the beef (R2 = 0.71). The level of correlation was comparable to inter-panel variation, which suggests a high degree of performance without over-fitting. The study clearly demonstrated that Raman spectroscopy could be used to assess the consumer perceived quality of a piece of meat.

References

1.         Beattie, R. J., et al., Meat Science 2004, 66, (4), 903-913.

2.         Beattie, J. R., et al., Meat Science 2008, 80, (4), 1205-1211.

 


   

Home | About Me | Publications | Resources | External Links | Sitemap