- Email: [email protected]
The Use of Near-Infrared Reflectance Spectroscopy in the Prediction of the Chemical Composition of Goose Fatty Liver
PROCESSING AND PRODUCTS The Use of Near-Infrared Reflectance Spectroscopy in the Prediction of the Chemical Composition of Goose Fatty Liver C. Molette,* P. Berzaghi,† A. Dalle Zotte,† H. Remignon,*,1 and R. Babile* *Ecole Nationale Supe´rieure Agronomique Toulouse—BP 107, 31326 Castanet Tolosan CEDEX, France; and †Department of Animal Science, Agripolis, 35020—Legnaro (PD), Italy (0.805) and DM (0.908) but were low for ash (0.151) and relatively low for protein content (0.255). For the major fatty acids, R2 ranged from 0.886 for palmitic acid to 0.988 for oleic acid. Oleic acid, the main fatty acid of the liver, and the stearic acid had higher R2 values than the less represented fatty acids. This study suggests that the NIRS technique can be used to predict lipid content and the fatty acid composition of goose fatty livers, but calibration must be built on a larger number of samples to generate accurate predictions.
(Key words: goose fatty liver, fatty acid composition, chemical composition, near-infrared reflectance spectroscopy) 2001 Poultry Science 80:1625–1629
INTRODUCTION In the 1950s, the USDA began studying the near-infrared (NIR) properties of optically opaque substances. The initial work in NIR spectroscopy involved methanol extracts or slurries to measure moisture in grain and seeds or used a thin layer of sample for measuring moisture and fat (BenGera and Norris, 1968). Today, the field of application of NIR measurements is tremendously wider. NIR analysis is used in agricultural and food products and also in the pharmaceutical field, petroleum and chemical industries, environmental applications, and cosmetic and textile manufacturing. The field of meat quality is not forgotten by NIR measurements. Indeed, almost all species have been studied using this method to determine the chemical composition of meats such as rabbit (Masoero et al., 1994), chicken (Sindic et al., 1993), lamb (Cozzolino et al., 1999), and Pekin duck (Ko¨hler et al., 1995). Predictions were performed on the main biochemical components of meat (DM, proteins, lipids, and ash). Only the determination of ash in all species resulted in low values of the coefficient of determination (R2 around 0.50). Results of the other parameters are generally poorer for lamb meat than for the other species. The best calibration was performed for chicken meat with an
2001 Poultry Science Association, Inc. Received for publication November 20, 2000. Accepted for publication May 26, 2001. 1 To whom correspondence should be addressed: [email protected].
R2 above 0.98 for the major components (DM, lipids, and proteins). Despite the fact that the content of these constituents (DM, lipids, and proteins) differs from one species to another, the various R2 are quite similar. The aim of the present investigation was to test the NIRS technique on a meat product with a special feature, because goose foie gras (fatty livers) comprise a majority of fat [between 40 and 60% of the fresh matter (Leprettre et al., 1998)]. Despite the fact that some NIRS measurements have already been done with fat products such as olive oil (Garrido-Varo et al., 1999), this is the first report, to our knowledge, on a high fat content animal product.
MATERIALS AND METHODS Animals The birds were bred under normal conditions from 1 d to 13 wk of age on an experimental farm in Coulaures, France. At this age, the geese entered a force-feeding period for 18 d. Briefly, the steatosis of the liver was induced by force-feeding the bird corn twice a day. The excessive consumption of corn provides an unbalanced diet that is rich in energy. This large consumption of carbohydrates induces the synthesis of fatty acids by the hepatic tissue. Because the maximum of exportation capacity of these fatty acids is rapidly reached, the liver becomes increas-
Abbreviation Key: NIRS = near infrared reflectance spectroscopy; 1-VR = coefficient of determination in the cross-validation.
1625
Downloaded from http://ps.oxfordjournals.org/ at North Dakota State University on May 20, 2015
ABSTRACT The use of near-infrared reflectance spectroscopy (NIRS) on a meat product is described in this report. The aim of the study was to develop calibration equations to predict the chemical composition of goose fatty liver (foie gras) with lipid contents greater than 40% of the fresh pate. Spectra of 52 foie gras samples were collected in the visible and NIR region (400 to 2,498 nm). Calibration equations were computed for DM, CP, lipids and fatty acids using modified partial least-squares regression. R2 values were high for the total lipid content
1626
MOLETTE ET AL.
TABLE 1. Chemical composition of the fatty livers (n = 52) Constituent1
Range (%)
Mean
SD
Protein Lipids Ash Dry matter
3.92–7.31 44.3–59.1 0.13–0.61 58.5–69.0
5.43 52.2 0.29 64.6
0.74 3.5 0.14 2.6
1
Percentage of frozen matter.
ingly fatty (for a review, see Cazeils, 2000; Hermier et al., 1999b). All animals were killed in a conventional slaughter house after a 12-h feed withdrawal. The samples of fatty liver were collected immediately after plucking, frozen in liquid nitrogen, and stored at −20 C until analyzed.
A representative sample of each fatty liver (n = 52) was ground as frozen, homogenized, and used for the determinations of the chemical composition of fatty livers. DM. The DM was determined after desiccation of the sample for 24 h in a drying oven at 103 C (JOCE, 1971a). Ash. Ash was determined according to the method of the JOCE (1971b) after the sample was charred in an muffle furnace at 550 C. Total Lipids. The lipids were extracted with a cold mixture of chloroform and methanol (two volumes of chloroform for one of methanol) according to the method described by Folch et al. (1957). CP. A wet mineralization (Kjeldhal type) was applied by using concentrated sulfuric acid with selenium added. The presence of nitrogen in the sample was then quantified colorimetrically (Verdouw et al., 1977). Protein content was expressed as total nitrogen multiplied by 6.25. Fatty Acid Composition. Fatty acid composition was determined after transmethylation (Morrison and Smith, 1964). Fatty acid methyl esters were analyzed using a gas chromatograph2 fitted with a 0.25 µm thick film of reticulated polyethyleneglycol phase and 30 m × 0.25 mm i.d. capillary column (Innovax),2 with nitrogen as the carrier gas. The temperatures of the injector and the detector were 250 and 300 C, respectively, and the column temperature was set from 200 to 250 C with a 5 C/min rise.
NIRS The fatty liver samples were preserved frozen at −20 C and then put in a 0 to 4 C cooler overnight before the NIRS examination. The measurements were carried out at room temperature in duplicate. The samples were fitted in ring cups (5 cm diameter) with a quartz glass and a polyester stopper. The near-infrared measurements were carried out on a FOSS NIRSystem 6500.3 Wavelengths from 400 to
2
Hewlett-Packard 5890 series II, Waldbronn, Germany. FOSS NIRSystem, Silver Spring, MD 20904. 4 WINISI 1.02, Infrasoft International Inc., Port Matilda, PA, 16870. 3
Statistical Analysis A commercial spectral analysis program4 was used to process the data and develop chemometric models. The method of the modified partial least-squares was used for all chemometric models. The best calibrations were always obtained with the whole spectrum of the NIR region and using the second derivative, a gap and smoothing factor calculated over four data points. Cross-validation was performed during model development, whereby one-sixth of the calibration samples at a time were temporarily removed from the calibration set. Performance statistics were accumulated for each group of removed samples. The optimization of the modified partial least-squares model for each tested constituents was obtained by selecting a number of partial least-squares factors that minimized the overall error between modeled and reference values (standard error of cross-validation). Because of the limited number of samples used, the number of partial least-squares factors was limited to five. Model performance was reported as the R2, the SE of calibration, and the R2 in the cross-validation (1-VR). The SE of calibration and of cross-validation represent the measurement of the uncertainty and were used to modulate the R2. The 1-VR corresponded to the proportion of reference method variation explained by the cross-validation predicted values.
RESULTS Chemical Analysis Table 1 reports the results of the biochemical composition of goose fatty livers used for calibrating the NIRS. The sum of ether extract, CP and ash do not equal DM because fatty livers contain 40.6 µmol of glucose and 57.7 µmol of glycogen per gram of fresh matter (Baron, 1998), which were not titrated. The liver weight was 769 ± 116 g. As reported in Table 1, the DM is the most important component and represents more than 60% of the fresh matter. The total lipid content of the fresh matter is also very high (on the average more than 50%) as was expected. Ash (<1%), protein (5%), and the other components (such as glycogen) represented only minor constituents of the foie. All these results agreed with those of Leprettre (1998), Salichon et al. (1994), and Hermier et al. (1999a).
Fatty Acids Table 2 summarizes the results obtained for the chemical analysis of the various fatty acids contents of the lipid extract. The sum of the fatty acids do not equal 100%
Downloaded from http://ps.oxfordjournals.org/ at North Dakota State University on May 20, 2015
Chemical Analysis
1,098 nm for the visible region and from 1,100 nm to 2,498 nm for the near-infrared region were measured. The detector readings were taken every second nanometer. The duplicate scans of each sample were examined for repeatability and then were averaged.
1627
NEAR-INFRARED REFLECTANCE SPECTROSCOPY AND GOOSE FATTY LIVERS TABLE 2. Percentages of the main fatty acids present in fatty livers (n = 52)
NIRS Calibration
Fatty Acids
The results obtained for the chemical composition by the NIRS system are reported in Table 3. The multiple coefficient of determination (R2) ranged from 0.151 for ash to 0.908 for DM. The R2 value was also high (0.805) for the total lipid content, whereas it remained relatively low for protein content (0.255). The R2 for the fatty acids ranged from 0.886 for palmitic acid to 0.988 for oleic acid (Table 4). The R2 values were higher for oleic and stearic acids than for the others. The 1-VR ranged from 0.165 for linoleic acid to 0.733 for stearic acid. Even if the 1-VR was very low for linoleic acid (0.165), the R2 remained high (0.914).
The calibration for lipids was quite accurate. The relationship between the predicted values in the cross-validation of stearic and oleic acids, respectively, and the reference values are shown on Figures 1 and 2. The diagonal line depicts the good agreement between predicted and reference values. As usual with birds, the unsaturated fatty acids represented a high percentage of total lipid; oleic and linoleic acids were 52.9 to 61.1% of the total fatty acids, which led to a high unsaturated:saturated fatty acid ratio (approximately 1.6). Lipids are synthesized from the carbohydrates in corn, and which explains the increase in the amount of saturated fatty acid and oleic acid that all birds synthesize easily. The low quantity of linoleic acid (less than 1% of the total fatty acid) is in agreement with previous studies (Baudonnet-Lenfant et al., 1991; Salichon et al., 1994; Hermier et al.; 1999a), which is surprising, given that the dietary uptake of this fatty acid is high. Indeed, linoleic acid represents 2% of the dry corn weight, which is about 5% of the metabolizable energy. It is hypothesized that preferential transport occurs to the peripheral adipose tissue. Our results were better than those found by Windham and Morrison (1998) in which the measurements of fatty acids were on beef neck. In the latter, the NIRS analysis was done on homogenized lean beef neck. The profile of
Fatty acid
Range (%)
Mean
SD
Myristic acid Palmitic acid Palmitoleic acid Stearic acid Oleic acid Linoleic acid
0.38–1.24 19.5–29.8 1.99–7.28 7.90–18.16 52.5–60.4 0.42–0.71
0.71 24.7 3.34 13.16 56.3 0.55
0.20 2.29 0.97 2.15 1.81 0.07
DISCUSSION Chemical Composition The R2 for DM, protein, and lipids were not as high as those obtained for meats: 0.93, 0.93, and 0.995, respectively, for rabbit meat (Masoero et al., 1994); 0.98, 0.99, 0.99, respectively, for chicken meat (Sindic et al., 1993), and 0.73 for intramuscular fat, 0.83 for CP, and 0.76 for moisture (Cozzolino et al., 1999) in lamb meat. In fatty liver, moisture and total lipids seem to be easy to calibrate by NIRS, which is probably due to their pre-
TABLE 3. Calibration data obtained by near-infrared reflectance spectroscopy for the determination of the chemical composition of fatty livers Constituent
n1
SEC1,2
R2
SECV1,2
1-VR1
Protein Lipids Ash Dry matter
51 48 52 49
0.619 1.557 0.124 0.725
0.255 0.805 0.151 0.908
0.679 1.618 0.133 0.960
0.095 0.792 0.027 0.841
1 n = number of samples; SEC = standard error of calibration; SECV = standard error of cross-validation; 1VR = coefficient of determination in the cross- validation. 2 Percentage of frozen matter.
Downloaded from http://ps.oxfordjournals.org/ at North Dakota State University on May 20, 2015
because only the fatty acids with more than 16 carbons were titrated, and those that represented less than 1% of the total fatty acids were considered as minor and were not reported. Oleic acid (C18:1) was the most important fatty acid because it represented more than 50% of the total fatty acids. Palmitic acid (C16:0) represented one-fourth of the total fatty acids, whereas stearic acid (C18:0) accounted for 13% of the total. The other fatty acids seemed to be of minor importance as they represented less than 5% of the total.
dominance in this product, as they represent nearly 90% of the total fresh matter. On the other hand, protein and ash represented less than 10% of the total fresh matter and had a much lower standard deviation, which could easily explain why they were poorly calibrated. Lipids are highly represented in the fatty liver product and may play a major role in a precise determination of the chemical composition with the NIRS system. Indeed, the O-H bond has a high molecular absorption around 1,900 nm. Garrido-Varo et al. (1999) have shown that, in olive oil, the NIRS system can explain 76% of the existing moisture of the product. Our product was similar to vegetable oil, as it is rich in lipids and relatively low in water. However, we obtained good accuracy for the calibration of the two major components. The poor value of R2 for ash could be explained by the fact that ash is not linked to a specific organic component of the analysis. Shenk and Westerhaus (1995) have demonstrated that minerals may only be detected in organic complexes or chelates by their effect on the hydrogen bond.
1628
MOLETTE ET AL. TABLE 4. Calibration data obtained by near-infrared reflectance spectroscopy for the determination of fatty acid profiles of fatty livers Fatty acid Myristic acid Palmitic acid Palmitoleic acid Stearic acid Oleic acid Linoleic acid
n
SEC1
R2
SECV1
1-VR
50 52 50 51 50 52
0.070 0.772 0.180 0.368 0.199 0.021
0.850 0.886 0.940 0.967 0.988 0.914
0.127 1.696 0.529 1.067 0.948 0.070
0.513 0.466 0.503 0.733 0.720 0.165
1 n = number of samples; SEC = standard error of calibration; SECV = standard error of cross-validation; 1VR = coefficient of determination in the cross-validation. 2 Percentage of frozen matter.
FIGURE 2. Plot of the analytical data (% of total fatty acids) and the values predicted (% of total fatty acids) in the cross-validation of the oleic acid content (n = 50). RSQ = r-squared; SEC = standard error of calibration; SECV = standard error of cross-validation; 1-VR = R2 in the cross-validation.
Downloaded from http://ps.oxfordjournals.org/ at North Dakota State University on May 20, 2015
FIGURE 1. Plot of the analytical data (% of total fatty acids) and the values predicted (% of total fatty acids) in the cross-validation of the stearic acid content (n = 51). RSQ = r-squared; SEC = standard error of calibration; SECV = standard error of cross-validation; 1-VR = R2 in the cross-validation.
NEAR-INFRARED REFLECTANCE SPECTROSCOPY AND GOOSE FATTY LIVERS
REFERENCES Baron, C., 1998. Comparaison de l’influence de deux syste`mes d’e´levage (intensif et traditionnel) et de deux techniques d’abattage (unique et fractionne´) sur les caracte´ristiques qualitatives du foie gras d’oie. Me´moire de DESS “Qualite´ des produits et se´curite´ alimentaire.” Baudonnet-Lenfant, C., A. Auvergne, and R. Babile, 1991. Influence de la dure´e du jeuˆne avant abattage et du poids avant la mise en gavage des canards de babarie sur la composition chimique he´patique. Ann. Zootech. 40:161–170. Ben-Gera, I., and K. H. Norris, 1968. Direct spectrophotometric determination of fat and moisture in meat products. J. Food Sci. 33:64–67. Cazeils, J. L., 2000. Caracte´risation de la composition lipidique des membranes plasmiques des he´patocytes de foies d’oies: relation avec le rendement technologique des foies gras. Ph.D. dissertation. Institut National Polytechnique, Toulouse, France. Cozzolino, D., I. Murray, J. R. Scaife, and R. Paterson, 1999. Lamb muscle identification by NIRS. pp 4–25 in: Proceedings of the 9th International Conference on Near-Infrared Spectroscopy. Conference Secretary, Verona, Italy.
De Pedro, E., A. Garrido, I. Barnes, M. Casillas, and I. Murray, 1992. Application of near infrared spectroscopy for quality control of Iberian pork industry. Pages 345–348 in: Near InfraRed Spectroscopy: Bridging the Gap Between Data Analysis and NIR Applications. K. I. Hildrum, T. Isaksoon, T. Naes, E. Horwood, ed. Chichester, U.K. Folch, J., M. Lees, and C. Sloane-Stanley, 1957. A simple method for the isolation and purification of total lipids from animal tissue. J. Biol Chem. 226:497–509. Garrido-Varo, A., C. Cobo, T. Sanchez-Pineda, R. Alcala, J. M. Horcas, and A. Jimenez, 1999. The feasibility of NIRS for olive oil quality control. p 4 in: Proceedings of the 9th International Conference on Near-Infrared Spectroscopy. Conference Secretary, Verona, Italy. Hermier, D., M. R. Salichon, G. Guy, R. Peresson, and J. Mourot, 1999a. Differential channelling of liver lipids in relation to susceptibility to hepatic steatosis in the goose. Poultry Sci. 78:1398–1406. Hermier, D., M. R. Salichon, G. Guy, R. Peresson, J. Mourot, and S. Lagarrigue, 1999b. La ste´atose he´patique des Palmipe`des gras; bases me´taboliques et sensibilite´ chimique. INRA Prod. Anim. 12:265–271. JOCE (Journal Officiel des Communaute´s Europe´ennes), 1971a. Dosage de l’humidite´. L279/8. Eur-Op News, Luxembourg, Luxembourg. JOCE (Journal Officiel des Communaute´s Europe´ennes), 1971b. Dosage des cendres brutes. L155/20. Eur-Op News, Luxembourg, Luxembourg. Ko¨hler, P., S. Wiederhold, and E. Kallweit, 1995. Near infrared transmission spectroscopy, a rapid method for evaluation of intramuscular fat and moisture content in Pekin ducks. Pages 368–372 in: Proceedings of the 10th European Symposium on Poultry, Halle, Germany. Leprettre, S., 1998. Incidence de facteurs peri-mortem sur les qualite´s technologiques et organoleptiques des foies gras d’oies. Etudes des de´fauts de couleurs. Ph.D. dissertation. Institut National Polytechnique, Toulouse, France. Leprettre, S., A. Auvergne, H. Manse, R. Babile, J. P. Dubois, and M. Candau, 1998. Incidence de la dure´e du jeuˆne sur la composition biochimique des foies gras d’oies et leur rendement a` la ste´rilisation. Sci. Aliment. 18:415–422. Morrison, W. R., and L. M. Smith, 1964. Preparation of fatty acid methyl esters and dimethylacetals from lipids with boron fluoride-methanol. J. Lipid Res. 5:600–608. Masoero, G., G. Xiccato, A. Dalle Zotte, R. Parigi-Bini, and G. Bergoglio, 1994. Analisi della carne di coniglio mediante spettroscopia NIR Zool. Nutr. Anim. 20:319–329. Reinhardt, T. C., C. Paul, and G. Robbelen, 1992. Pages 323–327 in: Making Light Work: Advances in Near Infrared Spectroscopy. I. Murray and I. A. Cowe, ed. VCH, Weinheim, Germany. Salichon, M. R., G. Guy, D. Rousselot, and J. C. Blum, 1994. Composition de trois types de foie gras: Oie, canard mulard et canard de Barbarie. Ann Zootech. 43:213–220. Shenk, J. S., and M. O. Westerhaus, 1995. Analysis of Agricultural and Food Products by Near-Infrared Reflectance Spectroscopy. NIRSystem, Port Matilda, PA. Shenk, J. S., J. R. Workman, and M. O. Westerhaus, 1992. Application of NIR spectroscopy to agricultural products. Pages 383– 481 in: Handbook of Near-Infrared Analysis. D. A. Burris and E. W. Ciurczak, ed. Marcel Dekker, New York, NY. Sindic, M., O. Chevalier, P. Dardenne, and C. Deroanne, 1993. Analyse de la viande de poulet par spectrome´trie proche infrarouge. Viandes Prod Carne´s. 14:95–98. Verdouw, H., C. J. A. Van Echteld, and E. M. J. Dekkers, 1977. Ammonium determination based on indophenol formation with sodium salicylate. Water Res. 12:399–402. Windham, W. R., and W. H. Morrison, 1998. Prediction of fatty acids content in beef neck lean by near infrared reflectance analysis. J. Near Infrared Spectrosc. 6:229–234.
Downloaded from http://ps.oxfordjournals.org/ at North Dakota State University on May 20, 2015
fatty acids used in their experiment was the same as ours; however, the calibration results were less homogeneous than ours. The R2 values ranged from 0.11 for linoleic acid to 0.89 for oleic acid. As in our study, the R2 of oleic acid had the highest accuracy, which may be due to its proportion in the product (56% in fatty liver; 40.8% in the fat portion of the beef neck). The NIRS calibration performed by Windham and Morrison (1998) consistently gave lower R2 values than in our study. Finally, the differences of NIR calibration that are observed are mainly due to the type of material analyzed. There is, on the one hand, homogenized meat, in which proteins represent about 20% of the frozen matter and lipids 9%, whereas on the other hand, in fatty liver, proteins represent about 5% of the frozen matter, and lipids represent more than 50%. Moreover, De Pedro et al. (1992) reported successful results of NIR transmittance calibration equations for the major fatty acids (palmitic, stearic, oleic, and linoleic acid) in Iberian swine carcasses. Others (De Pedro et al., 1992; Reinhardt et al., 1992) have shown that determining other minor fatty acids is quantitatively difficult. The failure to determine some individual fatty acids with a high R2 can be due to the similarities in their NIR absorption pattern, because different fatty acids have the same absorbing molecular group (–CH2–) and a narrow standard deviation of the data set. In conclusion, NIRS can be used to partially predict the chemical composition in goose fatty livers with good accuracy. The calibration could be improved especially for proteins. The DM and lipid portion were calibrated with high precision. The fatty acid content could be accurately determined. The results could also be improved by using more samples. The results of this study suggest that NIRS has potential in the determination of chemical composition of very fatty products such as fatty liver. Because of its speed of analysis and low operating costs, NIR spectroscopy can be used for estimating chemical composition in quality control processes.
1629
(Key words: goose fatty liver, fatty acid composition, chemical composition, near-infrared reflectance spectroscopy) 2001 Poultry Science 80:1625–1629
INTRODUCTION In the 1950s, the USDA began studying the near-infrared (NIR) properties of optically opaque substances. The initial work in NIR spectroscopy involved methanol extracts or slurries to measure moisture in grain and seeds or used a thin layer of sample for measuring moisture and fat (BenGera and Norris, 1968). Today, the field of application of NIR measurements is tremendously wider. NIR analysis is used in agricultural and food products and also in the pharmaceutical field, petroleum and chemical industries, environmental applications, and cosmetic and textile manufacturing. The field of meat quality is not forgotten by NIR measurements. Indeed, almost all species have been studied using this method to determine the chemical composition of meats such as rabbit (Masoero et al., 1994), chicken (Sindic et al., 1993), lamb (Cozzolino et al., 1999), and Pekin duck (Ko¨hler et al., 1995). Predictions were performed on the main biochemical components of meat (DM, proteins, lipids, and ash). Only the determination of ash in all species resulted in low values of the coefficient of determination (R2 around 0.50). Results of the other parameters are generally poorer for lamb meat than for the other species. The best calibration was performed for chicken meat with an
2001 Poultry Science Association, Inc. Received for publication November 20, 2000. Accepted for publication May 26, 2001. 1 To whom correspondence should be addressed: [email protected].
R2 above 0.98 for the major components (DM, lipids, and proteins). Despite the fact that the content of these constituents (DM, lipids, and proteins) differs from one species to another, the various R2 are quite similar. The aim of the present investigation was to test the NIRS technique on a meat product with a special feature, because goose foie gras (fatty livers) comprise a majority of fat [between 40 and 60% of the fresh matter (Leprettre et al., 1998)]. Despite the fact that some NIRS measurements have already been done with fat products such as olive oil (Garrido-Varo et al., 1999), this is the first report, to our knowledge, on a high fat content animal product.
MATERIALS AND METHODS Animals The birds were bred under normal conditions from 1 d to 13 wk of age on an experimental farm in Coulaures, France. At this age, the geese entered a force-feeding period for 18 d. Briefly, the steatosis of the liver was induced by force-feeding the bird corn twice a day. The excessive consumption of corn provides an unbalanced diet that is rich in energy. This large consumption of carbohydrates induces the synthesis of fatty acids by the hepatic tissue. Because the maximum of exportation capacity of these fatty acids is rapidly reached, the liver becomes increas-
Abbreviation Key: NIRS = near infrared reflectance spectroscopy; 1-VR = coefficient of determination in the cross-validation.
1625
Downloaded from http://ps.oxfordjournals.org/ at North Dakota State University on May 20, 2015
ABSTRACT The use of near-infrared reflectance spectroscopy (NIRS) on a meat product is described in this report. The aim of the study was to develop calibration equations to predict the chemical composition of goose fatty liver (foie gras) with lipid contents greater than 40% of the fresh pate. Spectra of 52 foie gras samples were collected in the visible and NIR region (400 to 2,498 nm). Calibration equations were computed for DM, CP, lipids and fatty acids using modified partial least-squares regression. R2 values were high for the total lipid content
1626
MOLETTE ET AL.
TABLE 1. Chemical composition of the fatty livers (n = 52) Constituent1
Range (%)
Mean
SD
Protein Lipids Ash Dry matter
3.92–7.31 44.3–59.1 0.13–0.61 58.5–69.0
5.43 52.2 0.29 64.6
0.74 3.5 0.14 2.6
1
Percentage of frozen matter.
ingly fatty (for a review, see Cazeils, 2000; Hermier et al., 1999b). All animals were killed in a conventional slaughter house after a 12-h feed withdrawal. The samples of fatty liver were collected immediately after plucking, frozen in liquid nitrogen, and stored at −20 C until analyzed.
A representative sample of each fatty liver (n = 52) was ground as frozen, homogenized, and used for the determinations of the chemical composition of fatty livers. DM. The DM was determined after desiccation of the sample for 24 h in a drying oven at 103 C (JOCE, 1971a). Ash. Ash was determined according to the method of the JOCE (1971b) after the sample was charred in an muffle furnace at 550 C. Total Lipids. The lipids were extracted with a cold mixture of chloroform and methanol (two volumes of chloroform for one of methanol) according to the method described by Folch et al. (1957). CP. A wet mineralization (Kjeldhal type) was applied by using concentrated sulfuric acid with selenium added. The presence of nitrogen in the sample was then quantified colorimetrically (Verdouw et al., 1977). Protein content was expressed as total nitrogen multiplied by 6.25. Fatty Acid Composition. Fatty acid composition was determined after transmethylation (Morrison and Smith, 1964). Fatty acid methyl esters were analyzed using a gas chromatograph2 fitted with a 0.25 µm thick film of reticulated polyethyleneglycol phase and 30 m × 0.25 mm i.d. capillary column (Innovax),2 with nitrogen as the carrier gas. The temperatures of the injector and the detector were 250 and 300 C, respectively, and the column temperature was set from 200 to 250 C with a 5 C/min rise.
NIRS The fatty liver samples were preserved frozen at −20 C and then put in a 0 to 4 C cooler overnight before the NIRS examination. The measurements were carried out at room temperature in duplicate. The samples were fitted in ring cups (5 cm diameter) with a quartz glass and a polyester stopper. The near-infrared measurements were carried out on a FOSS NIRSystem 6500.3 Wavelengths from 400 to
2
Hewlett-Packard 5890 series II, Waldbronn, Germany. FOSS NIRSystem, Silver Spring, MD 20904. 4 WINISI 1.02, Infrasoft International Inc., Port Matilda, PA, 16870. 3
Statistical Analysis A commercial spectral analysis program4 was used to process the data and develop chemometric models. The method of the modified partial least-squares was used for all chemometric models. The best calibrations were always obtained with the whole spectrum of the NIR region and using the second derivative, a gap and smoothing factor calculated over four data points. Cross-validation was performed during model development, whereby one-sixth of the calibration samples at a time were temporarily removed from the calibration set. Performance statistics were accumulated for each group of removed samples. The optimization of the modified partial least-squares model for each tested constituents was obtained by selecting a number of partial least-squares factors that minimized the overall error between modeled and reference values (standard error of cross-validation). Because of the limited number of samples used, the number of partial least-squares factors was limited to five. Model performance was reported as the R2, the SE of calibration, and the R2 in the cross-validation (1-VR). The SE of calibration and of cross-validation represent the measurement of the uncertainty and were used to modulate the R2. The 1-VR corresponded to the proportion of reference method variation explained by the cross-validation predicted values.
RESULTS Chemical Analysis Table 1 reports the results of the biochemical composition of goose fatty livers used for calibrating the NIRS. The sum of ether extract, CP and ash do not equal DM because fatty livers contain 40.6 µmol of glucose and 57.7 µmol of glycogen per gram of fresh matter (Baron, 1998), which were not titrated. The liver weight was 769 ± 116 g. As reported in Table 1, the DM is the most important component and represents more than 60% of the fresh matter. The total lipid content of the fresh matter is also very high (on the average more than 50%) as was expected. Ash (<1%), protein (5%), and the other components (such as glycogen) represented only minor constituents of the foie. All these results agreed with those of Leprettre (1998), Salichon et al. (1994), and Hermier et al. (1999a).
Fatty Acids Table 2 summarizes the results obtained for the chemical analysis of the various fatty acids contents of the lipid extract. The sum of the fatty acids do not equal 100%
Downloaded from http://ps.oxfordjournals.org/ at North Dakota State University on May 20, 2015
Chemical Analysis
1,098 nm for the visible region and from 1,100 nm to 2,498 nm for the near-infrared region were measured. The detector readings were taken every second nanometer. The duplicate scans of each sample were examined for repeatability and then were averaged.
1627
NEAR-INFRARED REFLECTANCE SPECTROSCOPY AND GOOSE FATTY LIVERS TABLE 2. Percentages of the main fatty acids present in fatty livers (n = 52)
NIRS Calibration
Fatty Acids
The results obtained for the chemical composition by the NIRS system are reported in Table 3. The multiple coefficient of determination (R2) ranged from 0.151 for ash to 0.908 for DM. The R2 value was also high (0.805) for the total lipid content, whereas it remained relatively low for protein content (0.255). The R2 for the fatty acids ranged from 0.886 for palmitic acid to 0.988 for oleic acid (Table 4). The R2 values were higher for oleic and stearic acids than for the others. The 1-VR ranged from 0.165 for linoleic acid to 0.733 for stearic acid. Even if the 1-VR was very low for linoleic acid (0.165), the R2 remained high (0.914).
The calibration for lipids was quite accurate. The relationship between the predicted values in the cross-validation of stearic and oleic acids, respectively, and the reference values are shown on Figures 1 and 2. The diagonal line depicts the good agreement between predicted and reference values. As usual with birds, the unsaturated fatty acids represented a high percentage of total lipid; oleic and linoleic acids were 52.9 to 61.1% of the total fatty acids, which led to a high unsaturated:saturated fatty acid ratio (approximately 1.6). Lipids are synthesized from the carbohydrates in corn, and which explains the increase in the amount of saturated fatty acid and oleic acid that all birds synthesize easily. The low quantity of linoleic acid (less than 1% of the total fatty acid) is in agreement with previous studies (Baudonnet-Lenfant et al., 1991; Salichon et al., 1994; Hermier et al.; 1999a), which is surprising, given that the dietary uptake of this fatty acid is high. Indeed, linoleic acid represents 2% of the dry corn weight, which is about 5% of the metabolizable energy. It is hypothesized that preferential transport occurs to the peripheral adipose tissue. Our results were better than those found by Windham and Morrison (1998) in which the measurements of fatty acids were on beef neck. In the latter, the NIRS analysis was done on homogenized lean beef neck. The profile of
Fatty acid
Range (%)
Mean
SD
Myristic acid Palmitic acid Palmitoleic acid Stearic acid Oleic acid Linoleic acid
0.38–1.24 19.5–29.8 1.99–7.28 7.90–18.16 52.5–60.4 0.42–0.71
0.71 24.7 3.34 13.16 56.3 0.55
0.20 2.29 0.97 2.15 1.81 0.07
DISCUSSION Chemical Composition The R2 for DM, protein, and lipids were not as high as those obtained for meats: 0.93, 0.93, and 0.995, respectively, for rabbit meat (Masoero et al., 1994); 0.98, 0.99, 0.99, respectively, for chicken meat (Sindic et al., 1993), and 0.73 for intramuscular fat, 0.83 for CP, and 0.76 for moisture (Cozzolino et al., 1999) in lamb meat. In fatty liver, moisture and total lipids seem to be easy to calibrate by NIRS, which is probably due to their pre-
TABLE 3. Calibration data obtained by near-infrared reflectance spectroscopy for the determination of the chemical composition of fatty livers Constituent
n1
SEC1,2
R2
SECV1,2
1-VR1
Protein Lipids Ash Dry matter
51 48 52 49
0.619 1.557 0.124 0.725
0.255 0.805 0.151 0.908
0.679 1.618 0.133 0.960
0.095 0.792 0.027 0.841
1 n = number of samples; SEC = standard error of calibration; SECV = standard error of cross-validation; 1VR = coefficient of determination in the cross- validation. 2 Percentage of frozen matter.
Downloaded from http://ps.oxfordjournals.org/ at North Dakota State University on May 20, 2015
because only the fatty acids with more than 16 carbons were titrated, and those that represented less than 1% of the total fatty acids were considered as minor and were not reported. Oleic acid (C18:1) was the most important fatty acid because it represented more than 50% of the total fatty acids. Palmitic acid (C16:0) represented one-fourth of the total fatty acids, whereas stearic acid (C18:0) accounted for 13% of the total. The other fatty acids seemed to be of minor importance as they represented less than 5% of the total.
dominance in this product, as they represent nearly 90% of the total fresh matter. On the other hand, protein and ash represented less than 10% of the total fresh matter and had a much lower standard deviation, which could easily explain why they were poorly calibrated. Lipids are highly represented in the fatty liver product and may play a major role in a precise determination of the chemical composition with the NIRS system. Indeed, the O-H bond has a high molecular absorption around 1,900 nm. Garrido-Varo et al. (1999) have shown that, in olive oil, the NIRS system can explain 76% of the existing moisture of the product. Our product was similar to vegetable oil, as it is rich in lipids and relatively low in water. However, we obtained good accuracy for the calibration of the two major components. The poor value of R2 for ash could be explained by the fact that ash is not linked to a specific organic component of the analysis. Shenk and Westerhaus (1995) have demonstrated that minerals may only be detected in organic complexes or chelates by their effect on the hydrogen bond.
1628
MOLETTE ET AL. TABLE 4. Calibration data obtained by near-infrared reflectance spectroscopy for the determination of fatty acid profiles of fatty livers Fatty acid Myristic acid Palmitic acid Palmitoleic acid Stearic acid Oleic acid Linoleic acid
n
SEC1
R2
SECV1
1-VR
50 52 50 51 50 52
0.070 0.772 0.180 0.368 0.199 0.021
0.850 0.886 0.940 0.967 0.988 0.914
0.127 1.696 0.529 1.067 0.948 0.070
0.513 0.466 0.503 0.733 0.720 0.165
1 n = number of samples; SEC = standard error of calibration; SECV = standard error of cross-validation; 1VR = coefficient of determination in the cross-validation. 2 Percentage of frozen matter.
FIGURE 2. Plot of the analytical data (% of total fatty acids) and the values predicted (% of total fatty acids) in the cross-validation of the oleic acid content (n = 50). RSQ = r-squared; SEC = standard error of calibration; SECV = standard error of cross-validation; 1-VR = R2 in the cross-validation.
Downloaded from http://ps.oxfordjournals.org/ at North Dakota State University on May 20, 2015
FIGURE 1. Plot of the analytical data (% of total fatty acids) and the values predicted (% of total fatty acids) in the cross-validation of the stearic acid content (n = 51). RSQ = r-squared; SEC = standard error of calibration; SECV = standard error of cross-validation; 1-VR = R2 in the cross-validation.
NEAR-INFRARED REFLECTANCE SPECTROSCOPY AND GOOSE FATTY LIVERS
REFERENCES Baron, C., 1998. Comparaison de l’influence de deux syste`mes d’e´levage (intensif et traditionnel) et de deux techniques d’abattage (unique et fractionne´) sur les caracte´ristiques qualitatives du foie gras d’oie. Me´moire de DESS “Qualite´ des produits et se´curite´ alimentaire.” Baudonnet-Lenfant, C., A. Auvergne, and R. Babile, 1991. Influence de la dure´e du jeuˆne avant abattage et du poids avant la mise en gavage des canards de babarie sur la composition chimique he´patique. Ann. Zootech. 40:161–170. Ben-Gera, I., and K. H. Norris, 1968. Direct spectrophotometric determination of fat and moisture in meat products. J. Food Sci. 33:64–67. Cazeils, J. L., 2000. Caracte´risation de la composition lipidique des membranes plasmiques des he´patocytes de foies d’oies: relation avec le rendement technologique des foies gras. Ph.D. dissertation. Institut National Polytechnique, Toulouse, France. Cozzolino, D., I. Murray, J. R. Scaife, and R. Paterson, 1999. Lamb muscle identification by NIRS. pp 4–25 in: Proceedings of the 9th International Conference on Near-Infrared Spectroscopy. Conference Secretary, Verona, Italy.
De Pedro, E., A. Garrido, I. Barnes, M. Casillas, and I. Murray, 1992. Application of near infrared spectroscopy for quality control of Iberian pork industry. Pages 345–348 in: Near InfraRed Spectroscopy: Bridging the Gap Between Data Analysis and NIR Applications. K. I. Hildrum, T. Isaksoon, T. Naes, E. Horwood, ed. Chichester, U.K. Folch, J., M. Lees, and C. Sloane-Stanley, 1957. A simple method for the isolation and purification of total lipids from animal tissue. J. Biol Chem. 226:497–509. Garrido-Varo, A., C. Cobo, T. Sanchez-Pineda, R. Alcala, J. M. Horcas, and A. Jimenez, 1999. The feasibility of NIRS for olive oil quality control. p 4 in: Proceedings of the 9th International Conference on Near-Infrared Spectroscopy. Conference Secretary, Verona, Italy. Hermier, D., M. R. Salichon, G. Guy, R. Peresson, and J. Mourot, 1999a. Differential channelling of liver lipids in relation to susceptibility to hepatic steatosis in the goose. Poultry Sci. 78:1398–1406. Hermier, D., M. R. Salichon, G. Guy, R. Peresson, J. Mourot, and S. Lagarrigue, 1999b. La ste´atose he´patique des Palmipe`des gras; bases me´taboliques et sensibilite´ chimique. INRA Prod. Anim. 12:265–271. JOCE (Journal Officiel des Communaute´s Europe´ennes), 1971a. Dosage de l’humidite´. L279/8. Eur-Op News, Luxembourg, Luxembourg. JOCE (Journal Officiel des Communaute´s Europe´ennes), 1971b. Dosage des cendres brutes. L155/20. Eur-Op News, Luxembourg, Luxembourg. Ko¨hler, P., S. Wiederhold, and E. Kallweit, 1995. Near infrared transmission spectroscopy, a rapid method for evaluation of intramuscular fat and moisture content in Pekin ducks. Pages 368–372 in: Proceedings of the 10th European Symposium on Poultry, Halle, Germany. Leprettre, S., 1998. Incidence de facteurs peri-mortem sur les qualite´s technologiques et organoleptiques des foies gras d’oies. Etudes des de´fauts de couleurs. Ph.D. dissertation. Institut National Polytechnique, Toulouse, France. Leprettre, S., A. Auvergne, H. Manse, R. Babile, J. P. Dubois, and M. Candau, 1998. Incidence de la dure´e du jeuˆne sur la composition biochimique des foies gras d’oies et leur rendement a` la ste´rilisation. Sci. Aliment. 18:415–422. Morrison, W. R., and L. M. Smith, 1964. Preparation of fatty acid methyl esters and dimethylacetals from lipids with boron fluoride-methanol. J. Lipid Res. 5:600–608. Masoero, G., G. Xiccato, A. Dalle Zotte, R. Parigi-Bini, and G. Bergoglio, 1994. Analisi della carne di coniglio mediante spettroscopia NIR Zool. Nutr. Anim. 20:319–329. Reinhardt, T. C., C. Paul, and G. Robbelen, 1992. Pages 323–327 in: Making Light Work: Advances in Near Infrared Spectroscopy. I. Murray and I. A. Cowe, ed. VCH, Weinheim, Germany. Salichon, M. R., G. Guy, D. Rousselot, and J. C. Blum, 1994. Composition de trois types de foie gras: Oie, canard mulard et canard de Barbarie. Ann Zootech. 43:213–220. Shenk, J. S., and M. O. Westerhaus, 1995. Analysis of Agricultural and Food Products by Near-Infrared Reflectance Spectroscopy. NIRSystem, Port Matilda, PA. Shenk, J. S., J. R. Workman, and M. O. Westerhaus, 1992. Application of NIR spectroscopy to agricultural products. Pages 383– 481 in: Handbook of Near-Infrared Analysis. D. A. Burris and E. W. Ciurczak, ed. Marcel Dekker, New York, NY. Sindic, M., O. Chevalier, P. Dardenne, and C. Deroanne, 1993. Analyse de la viande de poulet par spectrome´trie proche infrarouge. Viandes Prod Carne´s. 14:95–98. Verdouw, H., C. J. A. Van Echteld, and E. M. J. Dekkers, 1977. Ammonium determination based on indophenol formation with sodium salicylate. Water Res. 12:399–402. Windham, W. R., and W. H. Morrison, 1998. Prediction of fatty acids content in beef neck lean by near infrared reflectance analysis. J. Near Infrared Spectrosc. 6:229–234.
Downloaded from http://ps.oxfordjournals.org/ at North Dakota State University on May 20, 2015
fatty acids used in their experiment was the same as ours; however, the calibration results were less homogeneous than ours. The R2 values ranged from 0.11 for linoleic acid to 0.89 for oleic acid. As in our study, the R2 of oleic acid had the highest accuracy, which may be due to its proportion in the product (56% in fatty liver; 40.8% in the fat portion of the beef neck). The NIRS calibration performed by Windham and Morrison (1998) consistently gave lower R2 values than in our study. Finally, the differences of NIR calibration that are observed are mainly due to the type of material analyzed. There is, on the one hand, homogenized meat, in which proteins represent about 20% of the frozen matter and lipids 9%, whereas on the other hand, in fatty liver, proteins represent about 5% of the frozen matter, and lipids represent more than 50%. Moreover, De Pedro et al. (1992) reported successful results of NIR transmittance calibration equations for the major fatty acids (palmitic, stearic, oleic, and linoleic acid) in Iberian swine carcasses. Others (De Pedro et al., 1992; Reinhardt et al., 1992) have shown that determining other minor fatty acids is quantitatively difficult. The failure to determine some individual fatty acids with a high R2 can be due to the similarities in their NIR absorption pattern, because different fatty acids have the same absorbing molecular group (–CH2–) and a narrow standard deviation of the data set. In conclusion, NIRS can be used to partially predict the chemical composition in goose fatty livers with good accuracy. The calibration could be improved especially for proteins. The DM and lipid portion were calibrated with high precision. The fatty acid content could be accurately determined. The results could also be improved by using more samples. The results of this study suggest that NIRS has potential in the determination of chemical composition of very fatty products such as fatty liver. Because of its speed of analysis and low operating costs, NIR spectroscopy can be used for estimating chemical composition in quality control processes.
1629