- Email: [email protected]
Predicting frankfurters quality metrics using light backscatter
Journal of Food Engineering 143 (2014) 132–138
Contents lists available at ScienceDirect
Journal of Food Engineering journal homepage: www.elsevier.com/locate/jfoodeng
Predicting frankfurters quality metrics using light backscatter Gema Nieto a,b,c,⇑, Youling L. Xiong c, Fred Payne b, Manuel Castillo b,d a
Department of Food Technology, Nutrition and Food Science, Faculty of Veterinary Sciences, University of Murcia, Murcia, Spain Department of Biosystems and Agricultural Engineering, University of Kentucky, 128 C.E. Barnhart Building, Lexington, KY 40546-0276, USA c Department of Animal and Food Sciences, University of Kentucky, 206 W.P. Garrigus Building, KY 40546-0275, USA d Centre Especial de Recerca Planta de Tecnologia dels Aliments (CERPTA), Departament de Ciència Animal i dels Aliments, Facultat de Veterinària, Universitat Autònoma de Barcelona, 08193 Bellaterra, Spain b
a r t i c l e
i n f o
Article history: Received 10 February 2014 Received in revised form 18 June 2014 Accepted 19 June 2014 Available online 14 July 2014 Keywords: Fiber optic Light backscatter Sensor Prediction Frankfurters Emulsion stability
a b s t r a c t The objective of this study was to determine whether light backscatter response from fresh pork meat emulsions is correlated to final product stability indices, such as textural parameters, susceptibility of the emulsion to phase separation during cooking and lipid oxidation during subsequent storage. A specially designed fiber optic measurement system was used in combination with a miniature fiber optic spectrometer to determine the intensity of light backscatter within the wavelength range 300– 1100 nm at different radial distances (2, 2.5 and 3 mm) with respect to the light source in pork meat emulsions with two fat levels (15%, 30%) and two levels (0%, 2.5%) of the natural antioxidant hydrolyzed potato protein (HPP). Textural parameters (hardness, deformability, cohesiveness and breaking force), cooking loss, TBARS (1, 2, 3, and 7 days) and CIELAB color coordinates of emulsions were measured. The results showed that light backscatter response measured during meat emulsification has potential as an early predictor of emulsion stability during finely comminuted meat products manufacturing. Ó 2014 Elsevier Ltd. All rights reserved.
1. Introduction Meat emulsions such as frankfurters and bolognas are finely comminuted and cooked products composed of water, muscle proteins, fat particles, salt and small amounts of non-meat ingredients, where meat proteins serve as natural emulsifier. In this group of processed meat products, fat and protein concentration and their Abbreviations: a*, redness; FB, breaking force; b*, yellowness; C, cohesiveness; D, deformability; D, distance between optical fibers (D1 = 2 mm, D2 = 2.5 mm, D3 = 3 mm); D1Rat P1, ratio distance 1 peak 1; D1Rat P3, ratio distance 1 peak 3; L*, lightness; Lean, lean percentage; Cab, Chroma; CL, cooking loss; CLnor, normalized cooking loss; F1, first compression peak; F2, second compression peak; Hab, Hue; HPP, hydrolyzed potato protein; H, Hardness; IT, integration time; I, normalized light backscatter intensity; I3D1, normalized light backscatter intensity peak 3 distance 2 mm; I3D2, normalized light backscatter intensity peak 3 distance 2.5 mm; I3D3, normalized light backscatter intensity peak 3 distance 3 mm; N, Newton; RFL, fat/lean ratio; TBARS, thibarbituric acid-reactive substances; T, temperature; TPA, texture profile analysis; WF, weight of the final cooked emulsion; W0, initial weight of the emulsion; Ws, weigh of the sausage sample; k3D1, wavelength at the maximum intensity for peak 3 distance 2 mm; k3D2, wavelength at the maximum intensity for peak 3 distance 2.5 mm; k3D3, wavelength at the maximum intensity for peak 3 distance 3 mm; e, original height. ⇑ Corresponding author at: Department of Food Technology, Nutrition and Food Science, Faculty of Veterinary Sciences, University of Murcia, Murcia, Spain. Tel.: +34 868884798; fax: +34 868884147. E-mail address: [email protected] (G. Nieto). http://dx.doi.org/10.1016/j.jfoodeng.2014.06.027 0260-8774/Ó 2014 Elsevier Ltd. All rights reserved.
chemical interactions, especially those occurring during the emulsification process, exert a marked impact on the quality of the final product as they affect both emulsion stability (i.e., cooking losses) and lipid oxidation. Meat emulsion stability and cooking losses depends on fat stabilization by proteins. According to Barbut (1998), fat stabilization during chopping is due to the formation of a proteic film around the fat particles that allows retaining fat inside the protein matrix. During chopping, certain attractive forces (Jones, 1984) contribute to hold the raw materials together and create a homogeneous matrix structure (Allais et al., 2004). Excessive reduction of fat particles size and inadequate soluble protein extraction or fat to protein ratio could lead to reduced emulsification ability and increased fat oxidation tendency. Finely comminuted meat products are an integral part of diet in developed nations like the US (USDA, 2005) and have great economic importance. Frankfurters and bolognas are the most popular comminuted products in the US and account for the 25% of all sausages sold (NHDSC, 2008). Base on an average cooking loss (weight,%) of 2.64 (optimum chopping conditions), the estimated economic loss resulting from non optimum emulsion stability during the cooking process was estimated to range between 0.2 and 1.65 billion dollar per year. Improving process control and automation of the meat emulsification process will reduce the economical impact of emulsion breakdown in meat industry worldwide. Currently, there is a
G. Nieto et al. / Journal of Food Engineering 143 (2014) 132–138
lack of an effective on-line technology to select the optimum length of the chopping process during meat emulsification that results in maximum emulsion stability during the heat treatment. Control improvement will require the development of suitable sensor technologies to monitor the optimum level of emulsification that would maximize yield, quality and consistency of the finely comminuted meat products. The application of light scatter sensors for process control and optimization has been deeply studied in the dairy industry. Castillo et al. (2005) designed an optical sensor for monitoring milk coagulation and curd syneresis to predict the changes in moisture while the curd is draining in the cheese vat. In meat, various methods have been proposed over the years to short and classify meat that could assist on preventing cooking losses induced by inappropriate meat emulsification – i.e., to determine PSE meats (Bendall and Swatland, 1988), but there is a lack of an effective inline method to monitor meat emulsification. The use of a real-time meat emulsion stability sensor technology having the ability to determine the optimum chopping endpoint would significantly improve the current control over the chopping process preventing both under and over-chopping defects, which would result in evident final product yield, consistency and quality gain. According to Álvarez et al. (2007), monitoring meat emulsification would require a sensor capable of providing a representative signal from the beginning of the process, when the sample is extremely heterogeneous, to the end of the process, when the sample is relatively homogeneous. Our previous results suggest that an inline light scatter sensor might be able to provide useful information about the meat emulsification process. The goal of this work was to establish whether light scatter measured at several radial distances from the light source in fresh pork emulsions having a range of lipid oxidation and emulsion stability tendencies could be used to predict important final product stability indices such as textural parameters, susceptibility of the emulsion to phase separation during cooking and lipid oxidation during subsequent refrigerated storage.
133
Meter (CR-310 Minolta Camera Co., Ltd., Osaka, Japan) having a CIE standard ‘‘C’’ illuminant and 0° viewing angle geometry. Coordinates a* and b* were used, according to Hunt (1977), to calculate both Chroma, Cab, and Hue, Hab, values as follows:
C ab ¼
qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi 2 a2 þ b
Hab ¼ arctanðb =a Þ
ð1Þ ð2Þ
2.3. Cooking losses Once the chopping process was completed, cooking losses (CL) of each emulsion sample was measured in triplicate. CL was calculated from the weight of the final cooked emulsion (WF) and the initial weight (W0) of the sample before cooking as follows:
C L ¼ 100ð1 W F =W o Þ
ð3Þ
2.4. Texture profile analysis (TPA) The influence of HPP and fat concentration on textural properties of frankfurters was investigated by uniaxial compression tests using an Instron UTM Universal Testing Machine (Model 4301; Instron UTM Corp., Canton, MA, USA) as described by (Xiong et al., 1999). Cylindrical samples of 1.5 cm length were cut and compressed to 80% of its original height (strain, e = DL/L0 = 0.2, were L0 is the initial length of the cylinder) in a two cycle compression with 15 s delay between cycles. Hardness (H) of the sample was measured as the force (N) of the first compression peak (F1). The force of the second compression peak was designated as F2. The percent reduction in the compression force between the first and second compression peaks was defined as structure ‘‘Deformability’’ (D) and was calculated as D (%) = 100(F1 F2)/F1. Cohesiveness (C), as defined by Bourne (1978) (ratio of total areas between the first and second compression peaks) was estimated as (F2/F1)2 (dimensionless). Another set of samples was compressed to 20% of its original height (e = 0.8) to determine the breaking force (FB) (N).
2. Material and methods
2.5. Thibarbituric acid-reactive substances (TBARS)
2.1. Experimental design and meat emulsion manufacturing
TBARS were measured to evaluate lipid oxidation on days 0, 1, 3 and 7 of storage at 4 °C, according to the method described by Wang and Xiong (2005). The TBARS value, expressed as mg of malondialdehyde per kg of sausage sample, was calculated using the following equation:
A completely randomized factorial design with two factors and three replications was used. Two different HPP levels (0 –control– and 2.5% w/w) were tested within a range of emulsion breakdown and lipid oxidation tendencies that were induced by using two different levels of fat (15 and 30% w/w; i.e., fat/lean ratios –RFL– of 0.18 and 0.43, respectively). A total of 12 tests (N = nab = 322) were conducted with this design. A variety of final product quality indices (technological dependent variables) were determined to establish the degree of lipid oxidation (thiobarbituric acid-reactive substances or TBARS) and emulsion stability (cooking losses, hardness, ‘‘deformability’’, cohesiveness and breaking force) of the cooked meat emulsions. Reflection photometry parameters (CIELAB color coordinates) were collected from fresh emulsions. For further details regarding meat samples preparation, HPP preparation, meat emulsion manufacturing and measurement of mentioned final product quality indices see Nieto et al. (2009). Light scatter intensity (300–1100 nm) in fresh meat emulsions was measured at different radial distances (2, 2.5 and 3 mm: D1, D2 and D3, respectively) with respect to the light source as described below. 2.2. Color CIELAB color coordinates, L*, a*, and b*, were measured 1 h after the emulsion was prepared using a hand held tristimulus Chroma
TBARS ¼ 9:48ðA532 =W s Þ:
ð4Þ
where A532 was absorbance at 532 nm, Ws was the sausage sample weight (g), and 9.48 was a constant derived from the dilution factor and the molar extinction coefficient (1.52 103 M1 cm1) of the red TBA reaction product. 2.6. Light backscatter measurement of raw emulsions A dedicated laboratory optical sensor prototype was designed, built and tested (Álvarez et al., 2009) in the Food Engineering Lab (University of Kentucky) to measure light scatter of comminuted meats at different distances with the aim of identifying and detecting physical–chemical changes occurring during chopping that may be correlated to emulsion stability. This optical sensor prototype was designed to set the radial distance between the emitting and detecting optical fibers by means of a micrometer. Two small plastic probes were built and configured such that light scatter from the sample could be detected using a High-Resolution Fiber Optic Spectrometer (Model HR4000, Ocean Optics, Inc., Dunedin, FL, USA). The light source utilized was a tungsten halogen
G. Nieto et al. / Journal of Food Engineering 143 (2014) 132–138
(300–1100 nm) bulb (LS-1, Ocean Optics, Inc.). Fiber optic cables were manufactured using 600 lm diameter fibers (Spectran Specialty Optics, Avon, CN, USA). The terminating (i.e., measuring) ends of the two fibers were built into the plastic probes while the other two ends were connected, using an SMA connector, to the spectrometer and light source, respectively. The data acquisition system consisted of a PC connected by a USB port to the spectrometer and programmed for data acquisition with SprectraSuit Spectroscopy platform software (Ocean Optics, Inc.). Before each measurement, the terminating ends of the fibers were aligned vertically and horizontally to the same level. Emulsion samples were placed in a doublejacketed sample holder. The fiber tips were immersed into the emulsion sample up to a final depth of 12.7 mm from the surface of the sample. The temperature of the sample was controlled by means of connecting the sample holder to a water bath (Lauda Ecoline RE220. Brinkman Instruments Inc. NY. USA; ±0.01 °C of accuracy). The temperature of the emulsion was monitored using a Traceable Thermometer inserted into the emulsion. An opaque enclosure was used to isolate the sample from ambient light interference. Light scatter intensity of the samples was measured at the target radial distances, and at an integration time (IT) ranging from 19 to 60 s, where IT was the detector light exposure time. The light scattering spectral scans, I(k), were automatically processed by subtracting the respective dark spectral scans and dividing by the IT to give the light scattering normalized spectral scans, I(k) (bits s1). A number of optical parameters defined in Table 1 were obtained from the normalized spectral scans. Fig. 1 shows a typical pork emulsion normalized spectral scan as well as the corresponding optical parameters obtained. 2.7. Statistical analysis SAS software (SAS, version 9.1, 2002-2003, SAS Institute, Cary, NC, USA) was used to perform the statistical analysis of data. The best models for prediction of emulsion stability parameters were obtained using the GLM and the variable selection ‘‘maximum R2’’ procedures of SAS. 3. Result and discussion The best single-, two-, three- and four-variable models for predicting cooking losses, TBARS and texture parameters were selected using the SAS Maximum R2 procedure. The variables taken into consideration in the selection procedure were the color coordinates measured by a colorimeter and the optical parameters obtained from the light scatter spectral scan. 3.1. Prediction of cooking losses As shown in Table 2, the best single variable model (Model I) contained the independent variable L* and the intercept. This model explained 77% of the variability observed in cooking losses (CL). Model II showed that the second best predictive parameter of CL was b*. The inclusion of b* in Model I increased the coefficient
Table 1 Definition of optical parameters derived from the normalized light backscatter intensity spectral scan. Peak position
I
k
P1 (493–533 nm) P2 (560–584 nm) P3 (636–720 nm)
I1 I2 I3
k1 k2 k3
P1,2,3: Peak 1, 2 and 3; I, Normalized Light backscatter intensity; k, wavelength.
Normalized intensity I (bits s-1)
134
900
I3,λ3
800 700
Peak 3
600 500 I1,λ1
400
I2,λ2
300 200
Peak 1
Peak 2
100 0 400
450
500
550
600
650
700
Wavelength (nm) Fig. 1. Light backscatter intensity spectral scan. Measurement was made at distance of 2 mm from the emitting light source. For further details see explanation in the text and Table 1.
of determination (R2) from 0.77 to 0.95. The next best predictor of CL was the inclusion of k2. Inclusion of these parameters in Model II increased R2 from 0.95 to 0.96 while inclusion of I2, I3 on Model V only increased the R2 value slightly, from 0.95 to 0.97. According to these results, k2 and color parameters, L* and b* show potential for online prediction of CL. Note that color parameters L⁄ and b⁄ can be directly obtained from the light scatter spectral scan measured using our proposed sensor and SpectraSuite software (SpectraSuite v. 5.1., Ocean Optics B. V., Duiven Netherlands). This will allow the use of color parameters in predictive equations inline and without using a colorimeter. It should be also noted that the predictions proposed only used optical parameters and there is no need for including compositional parameters or independent experimental factors as predictors. A practical CL prediction model for industrial application would require at least two prediction terms, the optical parameters measured by the proposed sensor and the emulsion lightness, L*. A few studies have dealt with the use of on-line instrumental devices to minimize CL during chopping and/or optimize the texture of the cooked product. Barbut (1998) detected a correlation between L* and CL and proposed the use of a fiber optic probe as a quality control tool to monitor and predict CL in meat emulsions. Koolmes et al. (1993) proposed conductimetry measurements to determine the stability of meat emulsions while Álvarez et al. (2010) suggested that light extinction spectroscopy could provide information about emulsion stability. 3.2. Prediction of lipid oxidation Tables 3–5 show the predictive models for TBARS at day 1, 3 and 7, respectively. For the most part of TBARS values, they were predicted by the exclusive use of optical parameters (normalized intensities and wavelengths determined by the sensor), L* and b* (CIELAB coordinates: lightness, yellowness). Table 3 and Fig. 2a show the prediction of TBARS1 using model IV, contained the independent variable I2D3, I2D1, I2D3, k3D2 and the intercept. This model explained 98% of the variability observed in TBARS at day 1. In addition, Table 4 shows the prediction of TBARS3 using model IV. This model explained 91% of the variability observed in TBARS at day 7. As shown in Table 5, the best single variable model (model I) contained the independent variable I1 and the intercept. This model explained 72% of the variability observed in TBARS at day 7. Model II showed that the second best predictive parameter of TBARS7 was b*. The inclusion of b* in Model I increased the coefficient of determination (R2), from 0.72 to 0.78 (Model II). Other predictors of TBARS7 selected by the maximum R2 procedure of SAS were I2, lean (Model III), and I2, lean, k3 (Model IV). Finally, Model V allowed to predict TBARS7 with an R2 value of 0.94 (Fig. 2b) using
135
G. Nieto et al. / Journal of Food Engineering 143 (2014) 132–138 Table 2 Models for the prediction of cooking losses in meat emulsions. Model ***
I II*** III*** IV*** V***
b0 *
C L = b0 + L b1 CL = b0 + L*b1 + b*b2 CL = b0 + L*b1 + b*b2 + k2b3 CL = b0 + L*b1 + b*b2 + k2b3 + I2b4 CL = b0 + L*b1 + b*b2 + I1b3 + I3b4
b1 ***
***
39.3 26.8*** 147ns 168ns 22.6**
0.40 0.31*** 0.29*** 0.27** 0.27**
b2
b3
b4
R2
– 0.62*** 0.72*** 0.79** 0.59*
– – 0.30ns 0.34ns 0.03*
– – – 0.002ns 0.012*
0.77 0.95 0.96 0.96 0.97
k2, wavelength peak 2; I1, normalized light backscatter intensity peak 1; I2, normalized light backscatter intensity peak 2; I3, normalized light backscatter intensity peak 3; b0, b1, b2, b3, b4: regression coefficients; L*, Lightness; b*: Yellowness. * P < 0.05. ** P < 0.01. *** P < 0.001.
Table 3 Models for the prediction of lipid oxidation (TBARS) in meat emulsions at day 1. Model ***
I II*** III*** IV***
b0
b1 ns
TBARS1 = b0 + I2D3b1 TBARS1 = b0 + I2D3b1 + I2D1b2 TBARS1 = b0 + I2D3b1 + I2D1b2 + I3D1b3 TBARS1 = b0 + I2D3b1 + I2D1b2 + I2D3b3 + k3D2b4
***
0.03 0.05*** 0.05*** 0.05***
0.29 0.21ns 0.01ns 5.88ns
b2
b3
b4
R2
– 0.01*** 0.01*** 0.01***
– – 0.01** 0.01**
– – – 0.01ns
0.85 0.93 0.97 0.98
I2D3, normalized light backscatter intensity peak 2 distance 3; I2D1, normalized light backscatter intensity peak 2 distance 1; I3D1, normalized light backscatter intensity peak 3 distance 1; I2D3, normalized light backscatter intensity peak 2 distance 3; k3D2, wavelength peak 3 distance 2; b0, b1, b2, b3, b4: regression coefficients; D1,2,3: distances 1 (2 mm), 2 (2.5 mm) and 3 (3 mm). * P < 0.05. ** P < 0.01. *** P < 0.001.
Table 4 Models for the prediction of lipid oxidation (TBARS) in meat emulsions at day 3. Model ***
I II*** III*** IV***
b0 TBARS3 = b0 + I1b1 TBARS3 = b0 + I1b1 + k3b2 TBARS3 = b0 + I1b1 + k3b2 + D1RatP1b3 TBARS3 = b0 + I1b1 + k3b2 + D1RatP1b3 + D1RatP3b4
b1
1.55 46.3* 41.7* 51.7*
ns
***
0.04 0.06*** 0.06*** 0.07***
b2
b3
b4
R2
– 0.06* 0.06* 0.07*
– – 0.05ns 0.14ns
– – – 0.03ns
0.77 0.86 0.89 0.91
I1, normalized light backscatter intensity peak 1; k3, wavelength peak 3; P1,2,3: Peak 1, 2 and 3; I, Normalized Light backscatter intensity; D1Rat D1Rat P3, Ratio distance 1 peak 3; b0, b1, b2, b3, b4: regression coefficients; D1,2,3: distances 1 (2 mm), 2 (2.5 mm) and 3 (3 mm). P < 0.01. * P < 0.05. *** P < 0.001.
P1,
Ratio distance 1 peak 1;
**
Table 5 Models for the prediction of lipid oxidation (TBARS) in meat emulsions at day 7. Model ***
I II*** II*** III*** IV*** V***
b0 TBARS7 = b0 + I1b1 TBARS7 = b0 + I1b1 + b*b2 TBARS7 = b0 + b*b1 + Lean b2 TBARS7 = b0 + I2b1 + b*b2 + Lean b3 TBARS7 = b0 + I2b1 + b*b2 + Lean b3 + k3b4 TBARS7 = b0 + I2b1 + b*b2 + k3b3 + L*b4
b1 ns
0.97 10.3ns 8.85ns 8.40* 55.0ns 173**
***
0.06 0.05** 0.44** 0.02ns 0.04ns 0.04*
b2
b3
b4
R2
– 0.87ns 3.94*** 3.15** 3.69** 3.23***
– – – 0.35* 0.42** 0.11*
– – – – 0.60ns 0.77**
0.72 0.78 0.85 0.88 0.92 0.94
I1, normalized light backscatter intensity peak 1; I2, normalized light backscatter intensity peak 2; k3, wavelength peak 3; P1,2,3: Peak 1, 2 and 3; Lean, lean percentage; b0, b1, b2, b3, b4: regression coefficients; L*, Lightness; b*: Yellowness; D1,2,3: distances 1 (2 mm), 2 (2.5 mm) and 3 (3 mm); * P < 0.05. ** P < 0.01. *** P < 0.001.
only optical parameters (I2, b*, k3, L*). According to these results, I2, b*, k3, L*, show potential for online prediction of TBARS7. The proposed models could help to optimize the composition of the raw products in order to diminish the oxidative tendency of lipids during storage, which would in turn increase the maximum storage time. This behavior suggests a highly dependent correlation between functional properties of meat emulsions, such as fat/lean proportion, and specific peaks of the spectral scan. According to
Swatland (2002), in meat mixtures, lipid content is the major variable in these correlations and the strongest relationships are obtained with red or near infrared (NIR). These peak correlations are highly wavelength dependent and apparently contain information of functional properties that could be used to estimate normalized intensity of light backscatter signal. To our knowledge these are the first prediction of lipid oxidation in meat emulsion using light backscatter. So there is a difficulty to discuss our predictive equation with other authors.
136
G. Nieto et al. / Journal of Food Engineering 143 (2014) 132–138
3
Predicted TBARS1
Predicted TBARS7
Predicted TBARS 1 = 0.9965 TBARS1 R 2 = 0.9836
2,5 2 1,5 1
a
0,5
14
Predicted TBARS 7 = 0.9845 TBARS 7
12
R = 0.9356
2
10 8 6 4
b
2 0
0 0
0,5
1
1,5
2
2,5
0
3
2
4
6
1,10
Predicted C
Predicted H (N)
0,98 0,96
Predicted H (N) = 0.9996 H (N) R2 = 0.9713
1,20
1,00 0,90 0,80
c
0,70 0,8
1
0,88 0,86 0,84
d 0,8
0,85
0,9
0,95
C 14,9
2
R = 0.9703
Predicted BF (N)
Predicted D (%)
14
0,94 0,92 0,9
1,2
Predicted D (%) = 0.9919 D
6
12
Predicted (C) = 1.006 C R2 = 0.8915
H (N) 7
10
0,82 0,8
0,60 0,6
8
TBARS 7
TBARS 1
5 4 3 2
e
1
Predicted BF (N) = 0.9985 BF R² = 0.9941
12,9 10,9 8,9 6,9 4,9
f
2,9 0,9
0 0
1
2
3
4
5
6
0,9
7
5,9
10,9
D (%)
BF (N)
Fig. 2. Prediction of TBARS1 using model IV (a). Prediction of TBARS7 using model V (b). Prediction of Hardness (N) using model V (c). Prediction of Cohesiveness (dimensionless) using model V (d). Prediction of Damageability using model V (e). Prediction of breaking force (N) using model V (f).
Table 6 Models for the prediction of hardness in meat emulsions. Model ***
I II*** III*** IV*** V***
b0 *
H = b0 + L b1 H = b0 + Habb1 + I3D3b2 H = b0 + Habb1 + I3D3b2 + k1D1b3 H = b0 + Habb1 + I3D3b2 + k1D1b3 + k1D2b4 H = b0 + Habb1 + I3D3b2 + k1D1b3 + k2D3b4 + I3D2b5
b1 ***
2.75 2.67** 6.25ns 50.9* 66.9**
**
0.02 0.01** 0.01** 0.01*** 0.01***
b2
b3
b4
b5
R2
– 0.002*** 0.002*** 0.001*** 0.002***
– – 0.006ns 0.006ns 0.03**
– – – 0.12* 0.15**
– – – – 0.001*
0.55 0.83 0.85 0.93 0.97
I3D3, normalized light backscatter intensity peak 3 distance 3;; I3D2, normalized light backscatter intensity peak 3 distance 2, k1D1, wavelength peak 1 distance 3; k1D2, wavelength peak 1 distance 2; k2D3, wavelength peak 2 distance 3; b0, b1, b2, b3, b4: regression coefficients; D1,2,3: distances 1 (2 mm), 2 (2.5 mm) and 3 (3 mm); L*, Lightness; b*: Yellowness. Hab: Hue angle; * P < 0.05. ** P < 0.01. *** P < 0.001.
3.3. Prediction of texture Tables 6–9 show the predictive models for: hardness, cohesiveness, damageability and breaking force, respectively. All the texture parameters tested were predicted by the exclusive use of optical parameters and CIELAB coordinates. As shown in Table 6, the best single variable model (model I) contained the independent variable L* and an intercept. This model explained 55% of the variability observed in hardness. The inclusion of Hab, I3D3, k1D1, k2D3, I3D2 in Model I increased the coefficient
of determination (R2), from 0.55 to 0.97 (Model V). According to these results, hue, normalized intensity and wavelength, show potential for online prediction of hardness. As CL prediction, our predictions only used optical parameters and there is no need for including compositional parameters of experimental conditions as predictors. The result in Tables 7 and 8 for cohesiveness and damageability are in the same line that hardness. In both cases only optical parameters and color coordinates were used in the prediction models. Model V for cohesiveness had an R2 of 0.89 while Model 5 for damageability had an R2 of 0.84.
137
G. Nieto et al. / Journal of Food Engineering 143 (2014) 132–138 Table 7 Models for the prediction of cohesiveness in meat emulsions. Model ***
I II*** III*** IV*** V***
b0 C = b0 + k1D2b1 C = b0 + I3D1b1 + I3D3b2 C = b0 + I3D1b1 + I3D2b2 + k1D2b3 C = b0 + I3D1b1 + I3D2b2 + k1D2b3 + I2D3b4 C = b0 + I3D1b1 + k1D1b2 + Habb3 + I2D3b4 + I3D2b5
b1 ns
2.59 0.87*** 2.28ns 2.61ns 3.84**
*
0.07 0.001* 0.001* 0.001* 0.001**
b2
b3
b4
b5
R2
– 0.001* 0.001* 0.001ns 0.01**
– – 0.01ns 0.01* 0.002*
– – – 0.001ns 0.001*
– – – – 0.004***
0.36 0.54 0.70 0.75 0.89
k1D2, wavelength peak 1 distance 2; k1D2, wavelength peak 1 distance 2; I3D1; normalized light backscatter intensity peak 3 distance 1; I3D3, normalized light backscatter intensity peak 3 distance 3; I3D2, normalized light backscatter intensity peak 3 distance 2; I2D3, normalized light backscatter intensity peak 2 distance 3; b0, b1, b2, b3, b4: regression coefficients; D1,2,3: distances 1 (2 mm), 2 (2.5 mm) and 3 (3 mm); Hab: Hue angle; * P < 0.05. ** P < 0.01. *** P < 0.001.
Table 8 Models for the prediction of damageability in meat emulsions. Model ***
I II*** III*** IV*** V***
b0 D = b0 + k1D2b1 D = b0 + I3D1b1 + I3D2b2 D = b0 + I3D1b1 + I3D2b2 + I1D1b3 D = b0 + I3D1b1 + I3D2b2 + k3D3b3 + Cabb4 D = b0 + I3D1b1 + I3D2b2 + k3D3b3 + Cabb4 + I2D3b5
b1 *
179 6.58*** 9.43*** 15.0ns 15.9ns
ns
0.34 0.04** 0.05** 0.07** 0.07**
b2
b3
b4
b5
R2
– 0.06** 0.06** 0.11** 0.11**
– – 0.04ns 0.03ns 0.03*
– – – 0.35* 0.35*
– – – – 0.03ns
0.32 0.56 0.97 0.80 0.84
k1D2, wavelength peak 1 distance 2; k3D3, wavelength peak 3 distance 3; I3D1; normalized light backscatter intensity peak 3 distance 1; I3D2, normalized light backscatter intensity peak 3 distance 2; I3D3, normalized light backscatter intensity peak 3 distance 3; b0, b1, b2, b3, b4: regression coefficients; D1,2,3: distances 1 (2 mm), 2 (2.5 mm) and 3 (3 mm); L*, Cab: Chroma; * P < 0.05. ** P < 0.01. *** P < 0.001.
Table 9 Models for the prediction of breaking force in meat emulsions. Model I*** II*** III*** IV*** V***
BF = b0 + I3D3b1 BF = b0 + k1D2b1 + I3D2b2 BF = b0 + k1D2b1 + I3D1b2 + I3D2b3 BF = b0 + k1D2b1 + k2D1b2 + I3D1b3 + I3D2b4 BF = b0 + k1D2b1 + k1D3b2 + k2D1b3 + I3D1b4 + I3
D2b5
b0
b1
b2
b3
b4
b5
R2
4.93* 449** 330** 1049* 1190**
0.04*** 0.87** 0.64** 1.34*** 1.77***
– 0.03*** 3.15** 3.10** 0.35ns
– – 0.05** 0.05*** 3.40 **
– – –
– – – – 0.10***
0.81 0.90 0.96 0.98 0.99
0.11*** 0.05***
I3D3, normalized light backscatter intensity peak 3 distance 3; I3D2, normalized light backscatter intensity peak 3 distance 2; I3D1, normalized light backscatter intensity peak 3 distance 1; k1D2, wavelength peak 1 distance 2; k1D3; wavelength peak 1 distance 3; b0, b1, b2, b3, b4: regression coefficients; D1,2,3: distances 1 (2 mm), 2 (2.5 mm) and 3 (3 mm); * P < 0.05. ** P < 0.01. *** P < 0.001.
An excellent prediction was obtained for breaking force using 5 optical parameters and an intercept (Model V; R2 = 0.99; Table 9). However, as it can be observed in Table 9, Model 1 using only one predictor (I3) and an intercept had also a high R2 value, 0.81. Taking into account the results, the models explain 97% (Hardness: model V shown in Fig. 2c), 89% (Cohesiveness: model V shown in Fig. 2d), 97% (Damageability: model III shown in Fig. 2e) and 99% (Breaking force: model V shown in Fig. 2f) of the variability observed for the corresponding parameter. Our results are similar to those obtained by Allais et al. (2004) who suggested that prediction found a correlation between final texture parameters of meat emulsions and fluorescence spectroscopy measurements. However, these authors did not present any predictive equation. It is important to note the importance of proper texture for the commercialization of the product. Textural characteristics of food are important aspects of consumer acceptance. Based on consumers demand and the requirements of health authorities, the food research and the meat industry have developed new formulations by reducing or replacing the fat content (Mendoza et al., 2001). However, fermented sausages are among the meat products where
fat reduction is more complicated because fat confers important properties i.e. flavor, and texture (Mendoza et al., 2001). One of the most detrimental effects of fat reduction in fermented sausages is sensory texture (Muguerza et al., 2002). Therefore, the modification of the composition of fermented sausages by reducing its fat content can lead to negative changes in the sensory characteristics of these products and affect consumers’ acceptability. These results show that the use of optical sensor for the study of the light backscatter extinction in finely comminuted meat emulsions may have potential for predicting the texture of the final product. The proposed models help to optimize the texture of the meat emulsion in order to improve the final product.
4. Conclusions In this study, a new on-line optical sensor technology to select the optimum end-point of chopping that would be able to minimize cooking losses and maximize the yield was designed. It could be concluded that the use of a fiber optic light scatter sensor to monitor fresh meat emulsification during the chopping process
138
G. Nieto et al. / Journal of Food Engineering 143 (2014) 132–138
allows successfully predict, cooking losses, textural parameters and lipid oxidation tendency in the cooked product using optical parameters only. The proposed optical measurement system and obtained predictive equations strongly suggest that the development of an inline light scatter sensor for chopping time optimization is feasible. Further research and development efforts are needed to improve accuracy of the predictions and develop a dedicated inline fiber optic sensor. Acknowledgements The authors wish to thank the University of Murcia and the Department of Biosystems and Agricultural Engineering (University of Kentucky) for the economical support of this research and the Departments of Animal and Food Sciences and Biosystems and Agricultural Engineering (University of Kentucky) for allowing the use of their facilities. References Allais, I., Christophe, V., Pierre, A., Dufour, E., 2004. A rapid method based on frontface flurescence spectroscopy for the monitoring of the texture of meat emulsions and frankfurters. Meat Sci. 67, 219–229. Álvarez, D., Payne, F.A., Castillo, M., Xiong, Y.L., 2007. Application of backscatter Light extintion to determine the stability of beef emulsions with different fat/ lean ratios. ASABE, Annual International Meeting, Minneapolis, Minnesota, USA. Álvarez, D., Castillo, M., Payne, F.A., Xiong, Y.L., 2009. A novel fiber optic sensor to monitor beef emulsion stability using visible Light scattering. Meat Sci. 81, 456– 466. Álvarez, D., Castillo, M., Payne, F.A., Cox, Y.L., Xiong, Y.L., 2010. Application of light extinction to determine stability of beef emulsions. J. Food Eng. 96, 309–315.
Barbut, S., 1998. Use of a fiber optic probe to predict meat emulsion breakdown. Ital. J. Food Sci. 3, 253–259. Bendall, J.R., Swatland, H.J., 1988. A review of the relationships of pH with physical aspects of pork quality. Meat Sci. 24, 85–126. Bourne, M.C., 1978. Texture profile analysis. Food Technol. 41 (2), 163–178. Castillo, M., Payne, F.A., Shea, A., 2005. Development of a combined sensor technology for monitoring coagulation and syneresis operations in cheese making. In: Proceedings of ADSA annual meeting. Cincinnati, Ohio, USA. Hunt, R.W.G., 1977. The specification of colour appearance: I. Concepts and terms. Color Res. Appl. 2, 55–68. Jones, K.W., 1984. Protein-lipid interactions in processed meats. In: Proceedings Recipe Meat Conference, vol. 37, p. 52. Koolmes, P.A., Winjgaards, G., Tersteeg, M.H.G., Van Logtestijn, J.G., 1993. Changes in the microstructure of a comminuted meat system during heating. Food Struct. 12, 427–441. Mendoza, E., García, M.L., Casas, C., Selgas, M.D., 2001. Inulin as fat substitute in low fat, dry fermented sausages. Meat Sci. 57, 387–393. Muguerza, E., Fista, G., Ansorena, D., Astiasaran, I., Bloukas, J.G., 2002. Effect of fat level and partial replacement of pork backfat with olive oil on processing and quality characteristics of fermented sausages. Meat Sci. 61, 397–404. NHDSC, 2008. National Hot Dog and Sausage Council. Vital Hot Dogs Statistics. AMI Foundation, Washington, USA. Nieto, G., Castillo, M., Xiong, Y.L., Alvarez, D., Payne, F., Garrido, M.D., 2009. Antioxidants and emulsifying properties of alcalase-hydrolyzed potato proteins in meat emulsions with different fat concentration. Meat Sci. 83 (1), 24–30. SAS, 2002. Statistical Analysis System, Rel. 9.1. SAS Institute Inc., Cary, NC. Swatland, H.J., 2002. On-line monitoring of meat quality. In: Kerry, J., Kerry, J., Ledward, D. (Eds.), Meat Processing Improving Quality. Woodhead Publishing, pp. 193–216. USDA (United States Department of Agriculture), 2005. Meat Animal’s Production Disposition and Income 2004 Summary. Wang, L.L., Xiong, Y.L., 2005. Inhibition of lipid oxidation in cooked beef patties by hydrolyzed potato protein is related to its reducing and radical scavenging ability. J. Agricult. Food Chem. 53, 9186–9192. Xiong, Y.L., Noel, D.C., Moody, W.G., 1999. Textural and sensory properties of lowfat beef sausages with added water and polysaccharides as affected by pH and salt. J. Food Sci. 64, 550–554.
Contents lists available at ScienceDirect
Journal of Food Engineering journal homepage: www.elsevier.com/locate/jfoodeng
Predicting frankfurters quality metrics using light backscatter Gema Nieto a,b,c,⇑, Youling L. Xiong c, Fred Payne b, Manuel Castillo b,d a
Department of Food Technology, Nutrition and Food Science, Faculty of Veterinary Sciences, University of Murcia, Murcia, Spain Department of Biosystems and Agricultural Engineering, University of Kentucky, 128 C.E. Barnhart Building, Lexington, KY 40546-0276, USA c Department of Animal and Food Sciences, University of Kentucky, 206 W.P. Garrigus Building, KY 40546-0275, USA d Centre Especial de Recerca Planta de Tecnologia dels Aliments (CERPTA), Departament de Ciència Animal i dels Aliments, Facultat de Veterinària, Universitat Autònoma de Barcelona, 08193 Bellaterra, Spain b
a r t i c l e
i n f o
Article history: Received 10 February 2014 Received in revised form 18 June 2014 Accepted 19 June 2014 Available online 14 July 2014 Keywords: Fiber optic Light backscatter Sensor Prediction Frankfurters Emulsion stability
a b s t r a c t The objective of this study was to determine whether light backscatter response from fresh pork meat emulsions is correlated to final product stability indices, such as textural parameters, susceptibility of the emulsion to phase separation during cooking and lipid oxidation during subsequent storage. A specially designed fiber optic measurement system was used in combination with a miniature fiber optic spectrometer to determine the intensity of light backscatter within the wavelength range 300– 1100 nm at different radial distances (2, 2.5 and 3 mm) with respect to the light source in pork meat emulsions with two fat levels (15%, 30%) and two levels (0%, 2.5%) of the natural antioxidant hydrolyzed potato protein (HPP). Textural parameters (hardness, deformability, cohesiveness and breaking force), cooking loss, TBARS (1, 2, 3, and 7 days) and CIELAB color coordinates of emulsions were measured. The results showed that light backscatter response measured during meat emulsification has potential as an early predictor of emulsion stability during finely comminuted meat products manufacturing. Ó 2014 Elsevier Ltd. All rights reserved.
1. Introduction Meat emulsions such as frankfurters and bolognas are finely comminuted and cooked products composed of water, muscle proteins, fat particles, salt and small amounts of non-meat ingredients, where meat proteins serve as natural emulsifier. In this group of processed meat products, fat and protein concentration and their Abbreviations: a*, redness; FB, breaking force; b*, yellowness; C, cohesiveness; D, deformability; D, distance between optical fibers (D1 = 2 mm, D2 = 2.5 mm, D3 = 3 mm); D1Rat P1, ratio distance 1 peak 1; D1Rat P3, ratio distance 1 peak 3; L*, lightness; Lean, lean percentage; Cab, Chroma; CL, cooking loss; CLnor, normalized cooking loss; F1, first compression peak; F2, second compression peak; Hab, Hue; HPP, hydrolyzed potato protein; H, Hardness; IT, integration time; I, normalized light backscatter intensity; I3D1, normalized light backscatter intensity peak 3 distance 2 mm; I3D2, normalized light backscatter intensity peak 3 distance 2.5 mm; I3D3, normalized light backscatter intensity peak 3 distance 3 mm; N, Newton; RFL, fat/lean ratio; TBARS, thibarbituric acid-reactive substances; T, temperature; TPA, texture profile analysis; WF, weight of the final cooked emulsion; W0, initial weight of the emulsion; Ws, weigh of the sausage sample; k3D1, wavelength at the maximum intensity for peak 3 distance 2 mm; k3D2, wavelength at the maximum intensity for peak 3 distance 2.5 mm; k3D3, wavelength at the maximum intensity for peak 3 distance 3 mm; e, original height. ⇑ Corresponding author at: Department of Food Technology, Nutrition and Food Science, Faculty of Veterinary Sciences, University of Murcia, Murcia, Spain. Tel.: +34 868884798; fax: +34 868884147. E-mail address: [email protected] (G. Nieto). http://dx.doi.org/10.1016/j.jfoodeng.2014.06.027 0260-8774/Ó 2014 Elsevier Ltd. All rights reserved.
chemical interactions, especially those occurring during the emulsification process, exert a marked impact on the quality of the final product as they affect both emulsion stability (i.e., cooking losses) and lipid oxidation. Meat emulsion stability and cooking losses depends on fat stabilization by proteins. According to Barbut (1998), fat stabilization during chopping is due to the formation of a proteic film around the fat particles that allows retaining fat inside the protein matrix. During chopping, certain attractive forces (Jones, 1984) contribute to hold the raw materials together and create a homogeneous matrix structure (Allais et al., 2004). Excessive reduction of fat particles size and inadequate soluble protein extraction or fat to protein ratio could lead to reduced emulsification ability and increased fat oxidation tendency. Finely comminuted meat products are an integral part of diet in developed nations like the US (USDA, 2005) and have great economic importance. Frankfurters and bolognas are the most popular comminuted products in the US and account for the 25% of all sausages sold (NHDSC, 2008). Base on an average cooking loss (weight,%) of 2.64 (optimum chopping conditions), the estimated economic loss resulting from non optimum emulsion stability during the cooking process was estimated to range between 0.2 and 1.65 billion dollar per year. Improving process control and automation of the meat emulsification process will reduce the economical impact of emulsion breakdown in meat industry worldwide. Currently, there is a
G. Nieto et al. / Journal of Food Engineering 143 (2014) 132–138
lack of an effective on-line technology to select the optimum length of the chopping process during meat emulsification that results in maximum emulsion stability during the heat treatment. Control improvement will require the development of suitable sensor technologies to monitor the optimum level of emulsification that would maximize yield, quality and consistency of the finely comminuted meat products. The application of light scatter sensors for process control and optimization has been deeply studied in the dairy industry. Castillo et al. (2005) designed an optical sensor for monitoring milk coagulation and curd syneresis to predict the changes in moisture while the curd is draining in the cheese vat. In meat, various methods have been proposed over the years to short and classify meat that could assist on preventing cooking losses induced by inappropriate meat emulsification – i.e., to determine PSE meats (Bendall and Swatland, 1988), but there is a lack of an effective inline method to monitor meat emulsification. The use of a real-time meat emulsion stability sensor technology having the ability to determine the optimum chopping endpoint would significantly improve the current control over the chopping process preventing both under and over-chopping defects, which would result in evident final product yield, consistency and quality gain. According to Álvarez et al. (2007), monitoring meat emulsification would require a sensor capable of providing a representative signal from the beginning of the process, when the sample is extremely heterogeneous, to the end of the process, when the sample is relatively homogeneous. Our previous results suggest that an inline light scatter sensor might be able to provide useful information about the meat emulsification process. The goal of this work was to establish whether light scatter measured at several radial distances from the light source in fresh pork emulsions having a range of lipid oxidation and emulsion stability tendencies could be used to predict important final product stability indices such as textural parameters, susceptibility of the emulsion to phase separation during cooking and lipid oxidation during subsequent refrigerated storage.
133
Meter (CR-310 Minolta Camera Co., Ltd., Osaka, Japan) having a CIE standard ‘‘C’’ illuminant and 0° viewing angle geometry. Coordinates a* and b* were used, according to Hunt (1977), to calculate both Chroma, Cab, and Hue, Hab, values as follows:
C ab ¼
qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi 2 a2 þ b
Hab ¼ arctanðb =a Þ
ð1Þ ð2Þ
2.3. Cooking losses Once the chopping process was completed, cooking losses (CL) of each emulsion sample was measured in triplicate. CL was calculated from the weight of the final cooked emulsion (WF) and the initial weight (W0) of the sample before cooking as follows:
C L ¼ 100ð1 W F =W o Þ
ð3Þ
2.4. Texture profile analysis (TPA) The influence of HPP and fat concentration on textural properties of frankfurters was investigated by uniaxial compression tests using an Instron UTM Universal Testing Machine (Model 4301; Instron UTM Corp., Canton, MA, USA) as described by (Xiong et al., 1999). Cylindrical samples of 1.5 cm length were cut and compressed to 80% of its original height (strain, e = DL/L0 = 0.2, were L0 is the initial length of the cylinder) in a two cycle compression with 15 s delay between cycles. Hardness (H) of the sample was measured as the force (N) of the first compression peak (F1). The force of the second compression peak was designated as F2. The percent reduction in the compression force between the first and second compression peaks was defined as structure ‘‘Deformability’’ (D) and was calculated as D (%) = 100(F1 F2)/F1. Cohesiveness (C), as defined by Bourne (1978) (ratio of total areas between the first and second compression peaks) was estimated as (F2/F1)2 (dimensionless). Another set of samples was compressed to 20% of its original height (e = 0.8) to determine the breaking force (FB) (N).
2. Material and methods
2.5. Thibarbituric acid-reactive substances (TBARS)
2.1. Experimental design and meat emulsion manufacturing
TBARS were measured to evaluate lipid oxidation on days 0, 1, 3 and 7 of storage at 4 °C, according to the method described by Wang and Xiong (2005). The TBARS value, expressed as mg of malondialdehyde per kg of sausage sample, was calculated using the following equation:
A completely randomized factorial design with two factors and three replications was used. Two different HPP levels (0 –control– and 2.5% w/w) were tested within a range of emulsion breakdown and lipid oxidation tendencies that were induced by using two different levels of fat (15 and 30% w/w; i.e., fat/lean ratios –RFL– of 0.18 and 0.43, respectively). A total of 12 tests (N = nab = 322) were conducted with this design. A variety of final product quality indices (technological dependent variables) were determined to establish the degree of lipid oxidation (thiobarbituric acid-reactive substances or TBARS) and emulsion stability (cooking losses, hardness, ‘‘deformability’’, cohesiveness and breaking force) of the cooked meat emulsions. Reflection photometry parameters (CIELAB color coordinates) were collected from fresh emulsions. For further details regarding meat samples preparation, HPP preparation, meat emulsion manufacturing and measurement of mentioned final product quality indices see Nieto et al. (2009). Light scatter intensity (300–1100 nm) in fresh meat emulsions was measured at different radial distances (2, 2.5 and 3 mm: D1, D2 and D3, respectively) with respect to the light source as described below. 2.2. Color CIELAB color coordinates, L*, a*, and b*, were measured 1 h after the emulsion was prepared using a hand held tristimulus Chroma
TBARS ¼ 9:48ðA532 =W s Þ:
ð4Þ
where A532 was absorbance at 532 nm, Ws was the sausage sample weight (g), and 9.48 was a constant derived from the dilution factor and the molar extinction coefficient (1.52 103 M1 cm1) of the red TBA reaction product. 2.6. Light backscatter measurement of raw emulsions A dedicated laboratory optical sensor prototype was designed, built and tested (Álvarez et al., 2009) in the Food Engineering Lab (University of Kentucky) to measure light scatter of comminuted meats at different distances with the aim of identifying and detecting physical–chemical changes occurring during chopping that may be correlated to emulsion stability. This optical sensor prototype was designed to set the radial distance between the emitting and detecting optical fibers by means of a micrometer. Two small plastic probes were built and configured such that light scatter from the sample could be detected using a High-Resolution Fiber Optic Spectrometer (Model HR4000, Ocean Optics, Inc., Dunedin, FL, USA). The light source utilized was a tungsten halogen
G. Nieto et al. / Journal of Food Engineering 143 (2014) 132–138
(300–1100 nm) bulb (LS-1, Ocean Optics, Inc.). Fiber optic cables were manufactured using 600 lm diameter fibers (Spectran Specialty Optics, Avon, CN, USA). The terminating (i.e., measuring) ends of the two fibers were built into the plastic probes while the other two ends were connected, using an SMA connector, to the spectrometer and light source, respectively. The data acquisition system consisted of a PC connected by a USB port to the spectrometer and programmed for data acquisition with SprectraSuit Spectroscopy platform software (Ocean Optics, Inc.). Before each measurement, the terminating ends of the fibers were aligned vertically and horizontally to the same level. Emulsion samples were placed in a doublejacketed sample holder. The fiber tips were immersed into the emulsion sample up to a final depth of 12.7 mm from the surface of the sample. The temperature of the sample was controlled by means of connecting the sample holder to a water bath (Lauda Ecoline RE220. Brinkman Instruments Inc. NY. USA; ±0.01 °C of accuracy). The temperature of the emulsion was monitored using a Traceable Thermometer inserted into the emulsion. An opaque enclosure was used to isolate the sample from ambient light interference. Light scatter intensity of the samples was measured at the target radial distances, and at an integration time (IT) ranging from 19 to 60 s, where IT was the detector light exposure time. The light scattering spectral scans, I(k), were automatically processed by subtracting the respective dark spectral scans and dividing by the IT to give the light scattering normalized spectral scans, I(k) (bits s1). A number of optical parameters defined in Table 1 were obtained from the normalized spectral scans. Fig. 1 shows a typical pork emulsion normalized spectral scan as well as the corresponding optical parameters obtained. 2.7. Statistical analysis SAS software (SAS, version 9.1, 2002-2003, SAS Institute, Cary, NC, USA) was used to perform the statistical analysis of data. The best models for prediction of emulsion stability parameters were obtained using the GLM and the variable selection ‘‘maximum R2’’ procedures of SAS. 3. Result and discussion The best single-, two-, three- and four-variable models for predicting cooking losses, TBARS and texture parameters were selected using the SAS Maximum R2 procedure. The variables taken into consideration in the selection procedure were the color coordinates measured by a colorimeter and the optical parameters obtained from the light scatter spectral scan. 3.1. Prediction of cooking losses As shown in Table 2, the best single variable model (Model I) contained the independent variable L* and the intercept. This model explained 77% of the variability observed in cooking losses (CL). Model II showed that the second best predictive parameter of CL was b*. The inclusion of b* in Model I increased the coefficient
Table 1 Definition of optical parameters derived from the normalized light backscatter intensity spectral scan. Peak position
I
k
P1 (493–533 nm) P2 (560–584 nm) P3 (636–720 nm)
I1 I2 I3
k1 k2 k3
P1,2,3: Peak 1, 2 and 3; I, Normalized Light backscatter intensity; k, wavelength.
Normalized intensity I (bits s-1)
134
900
I3,λ3
800 700
Peak 3
600 500 I1,λ1
400
I2,λ2
300 200
Peak 1
Peak 2
100 0 400
450
500
550
600
650
700
Wavelength (nm) Fig. 1. Light backscatter intensity spectral scan. Measurement was made at distance of 2 mm from the emitting light source. For further details see explanation in the text and Table 1.
of determination (R2) from 0.77 to 0.95. The next best predictor of CL was the inclusion of k2. Inclusion of these parameters in Model II increased R2 from 0.95 to 0.96 while inclusion of I2, I3 on Model V only increased the R2 value slightly, from 0.95 to 0.97. According to these results, k2 and color parameters, L* and b* show potential for online prediction of CL. Note that color parameters L⁄ and b⁄ can be directly obtained from the light scatter spectral scan measured using our proposed sensor and SpectraSuite software (SpectraSuite v. 5.1., Ocean Optics B. V., Duiven Netherlands). This will allow the use of color parameters in predictive equations inline and without using a colorimeter. It should be also noted that the predictions proposed only used optical parameters and there is no need for including compositional parameters or independent experimental factors as predictors. A practical CL prediction model for industrial application would require at least two prediction terms, the optical parameters measured by the proposed sensor and the emulsion lightness, L*. A few studies have dealt with the use of on-line instrumental devices to minimize CL during chopping and/or optimize the texture of the cooked product. Barbut (1998) detected a correlation between L* and CL and proposed the use of a fiber optic probe as a quality control tool to monitor and predict CL in meat emulsions. Koolmes et al. (1993) proposed conductimetry measurements to determine the stability of meat emulsions while Álvarez et al. (2010) suggested that light extinction spectroscopy could provide information about emulsion stability. 3.2. Prediction of lipid oxidation Tables 3–5 show the predictive models for TBARS at day 1, 3 and 7, respectively. For the most part of TBARS values, they were predicted by the exclusive use of optical parameters (normalized intensities and wavelengths determined by the sensor), L* and b* (CIELAB coordinates: lightness, yellowness). Table 3 and Fig. 2a show the prediction of TBARS1 using model IV, contained the independent variable I2D3, I2D1, I2D3, k3D2 and the intercept. This model explained 98% of the variability observed in TBARS at day 1. In addition, Table 4 shows the prediction of TBARS3 using model IV. This model explained 91% of the variability observed in TBARS at day 7. As shown in Table 5, the best single variable model (model I) contained the independent variable I1 and the intercept. This model explained 72% of the variability observed in TBARS at day 7. Model II showed that the second best predictive parameter of TBARS7 was b*. The inclusion of b* in Model I increased the coefficient of determination (R2), from 0.72 to 0.78 (Model II). Other predictors of TBARS7 selected by the maximum R2 procedure of SAS were I2, lean (Model III), and I2, lean, k3 (Model IV). Finally, Model V allowed to predict TBARS7 with an R2 value of 0.94 (Fig. 2b) using
135
G. Nieto et al. / Journal of Food Engineering 143 (2014) 132–138 Table 2 Models for the prediction of cooking losses in meat emulsions. Model ***
I II*** III*** IV*** V***
b0 *
C L = b0 + L b1 CL = b0 + L*b1 + b*b2 CL = b0 + L*b1 + b*b2 + k2b3 CL = b0 + L*b1 + b*b2 + k2b3 + I2b4 CL = b0 + L*b1 + b*b2 + I1b3 + I3b4
b1 ***
***
39.3 26.8*** 147ns 168ns 22.6**
0.40 0.31*** 0.29*** 0.27** 0.27**
b2
b3
b4
R2
– 0.62*** 0.72*** 0.79** 0.59*
– – 0.30ns 0.34ns 0.03*
– – – 0.002ns 0.012*
0.77 0.95 0.96 0.96 0.97
k2, wavelength peak 2; I1, normalized light backscatter intensity peak 1; I2, normalized light backscatter intensity peak 2; I3, normalized light backscatter intensity peak 3; b0, b1, b2, b3, b4: regression coefficients; L*, Lightness; b*: Yellowness. * P < 0.05. ** P < 0.01. *** P < 0.001.
Table 3 Models for the prediction of lipid oxidation (TBARS) in meat emulsions at day 1. Model ***
I II*** III*** IV***
b0
b1 ns
TBARS1 = b0 + I2D3b1 TBARS1 = b0 + I2D3b1 + I2D1b2 TBARS1 = b0 + I2D3b1 + I2D1b2 + I3D1b3 TBARS1 = b0 + I2D3b1 + I2D1b2 + I2D3b3 + k3D2b4
***
0.03 0.05*** 0.05*** 0.05***
0.29 0.21ns 0.01ns 5.88ns
b2
b3
b4
R2
– 0.01*** 0.01*** 0.01***
– – 0.01** 0.01**
– – – 0.01ns
0.85 0.93 0.97 0.98
I2D3, normalized light backscatter intensity peak 2 distance 3; I2D1, normalized light backscatter intensity peak 2 distance 1; I3D1, normalized light backscatter intensity peak 3 distance 1; I2D3, normalized light backscatter intensity peak 2 distance 3; k3D2, wavelength peak 3 distance 2; b0, b1, b2, b3, b4: regression coefficients; D1,2,3: distances 1 (2 mm), 2 (2.5 mm) and 3 (3 mm). * P < 0.05. ** P < 0.01. *** P < 0.001.
Table 4 Models for the prediction of lipid oxidation (TBARS) in meat emulsions at day 3. Model ***
I II*** III*** IV***
b0 TBARS3 = b0 + I1b1 TBARS3 = b0 + I1b1 + k3b2 TBARS3 = b0 + I1b1 + k3b2 + D1RatP1b3 TBARS3 = b0 + I1b1 + k3b2 + D1RatP1b3 + D1RatP3b4
b1
1.55 46.3* 41.7* 51.7*
ns
***
0.04 0.06*** 0.06*** 0.07***
b2
b3
b4
R2
– 0.06* 0.06* 0.07*
– – 0.05ns 0.14ns
– – – 0.03ns
0.77 0.86 0.89 0.91
I1, normalized light backscatter intensity peak 1; k3, wavelength peak 3; P1,2,3: Peak 1, 2 and 3; I, Normalized Light backscatter intensity; D1Rat D1Rat P3, Ratio distance 1 peak 3; b0, b1, b2, b3, b4: regression coefficients; D1,2,3: distances 1 (2 mm), 2 (2.5 mm) and 3 (3 mm). P < 0.01. * P < 0.05. *** P < 0.001.
P1,
Ratio distance 1 peak 1;
**
Table 5 Models for the prediction of lipid oxidation (TBARS) in meat emulsions at day 7. Model ***
I II*** II*** III*** IV*** V***
b0 TBARS7 = b0 + I1b1 TBARS7 = b0 + I1b1 + b*b2 TBARS7 = b0 + b*b1 + Lean b2 TBARS7 = b0 + I2b1 + b*b2 + Lean b3 TBARS7 = b0 + I2b1 + b*b2 + Lean b3 + k3b4 TBARS7 = b0 + I2b1 + b*b2 + k3b3 + L*b4
b1 ns
0.97 10.3ns 8.85ns 8.40* 55.0ns 173**
***
0.06 0.05** 0.44** 0.02ns 0.04ns 0.04*
b2
b3
b4
R2
– 0.87ns 3.94*** 3.15** 3.69** 3.23***
– – – 0.35* 0.42** 0.11*
– – – – 0.60ns 0.77**
0.72 0.78 0.85 0.88 0.92 0.94
I1, normalized light backscatter intensity peak 1; I2, normalized light backscatter intensity peak 2; k3, wavelength peak 3; P1,2,3: Peak 1, 2 and 3; Lean, lean percentage; b0, b1, b2, b3, b4: regression coefficients; L*, Lightness; b*: Yellowness; D1,2,3: distances 1 (2 mm), 2 (2.5 mm) and 3 (3 mm); * P < 0.05. ** P < 0.01. *** P < 0.001.
only optical parameters (I2, b*, k3, L*). According to these results, I2, b*, k3, L*, show potential for online prediction of TBARS7. The proposed models could help to optimize the composition of the raw products in order to diminish the oxidative tendency of lipids during storage, which would in turn increase the maximum storage time. This behavior suggests a highly dependent correlation between functional properties of meat emulsions, such as fat/lean proportion, and specific peaks of the spectral scan. According to
Swatland (2002), in meat mixtures, lipid content is the major variable in these correlations and the strongest relationships are obtained with red or near infrared (NIR). These peak correlations are highly wavelength dependent and apparently contain information of functional properties that could be used to estimate normalized intensity of light backscatter signal. To our knowledge these are the first prediction of lipid oxidation in meat emulsion using light backscatter. So there is a difficulty to discuss our predictive equation with other authors.
136
G. Nieto et al. / Journal of Food Engineering 143 (2014) 132–138
3
Predicted TBARS1
Predicted TBARS7
Predicted TBARS 1 = 0.9965 TBARS1 R 2 = 0.9836
2,5 2 1,5 1
a
0,5
14
Predicted TBARS 7 = 0.9845 TBARS 7
12
R = 0.9356
2
10 8 6 4
b
2 0
0 0
0,5
1
1,5
2
2,5
0
3
2
4
6
1,10
Predicted C
Predicted H (N)
0,98 0,96
Predicted H (N) = 0.9996 H (N) R2 = 0.9713
1,20
1,00 0,90 0,80
c
0,70 0,8
1
0,88 0,86 0,84
d 0,8
0,85
0,9
0,95
C 14,9
2
R = 0.9703
Predicted BF (N)
Predicted D (%)
14
0,94 0,92 0,9
1,2
Predicted D (%) = 0.9919 D
6
12
Predicted (C) = 1.006 C R2 = 0.8915
H (N) 7
10
0,82 0,8
0,60 0,6
8
TBARS 7
TBARS 1
5 4 3 2
e
1
Predicted BF (N) = 0.9985 BF R² = 0.9941
12,9 10,9 8,9 6,9 4,9
f
2,9 0,9
0 0
1
2
3
4
5
6
0,9
7
5,9
10,9
D (%)
BF (N)
Fig. 2. Prediction of TBARS1 using model IV (a). Prediction of TBARS7 using model V (b). Prediction of Hardness (N) using model V (c). Prediction of Cohesiveness (dimensionless) using model V (d). Prediction of Damageability using model V (e). Prediction of breaking force (N) using model V (f).
Table 6 Models for the prediction of hardness in meat emulsions. Model ***
I II*** III*** IV*** V***
b0 *
H = b0 + L b1 H = b0 + Habb1 + I3D3b2 H = b0 + Habb1 + I3D3b2 + k1D1b3 H = b0 + Habb1 + I3D3b2 + k1D1b3 + k1D2b4 H = b0 + Habb1 + I3D3b2 + k1D1b3 + k2D3b4 + I3D2b5
b1 ***
2.75 2.67** 6.25ns 50.9* 66.9**
**
0.02 0.01** 0.01** 0.01*** 0.01***
b2
b3
b4
b5
R2
– 0.002*** 0.002*** 0.001*** 0.002***
– – 0.006ns 0.006ns 0.03**
– – – 0.12* 0.15**
– – – – 0.001*
0.55 0.83 0.85 0.93 0.97
I3D3, normalized light backscatter intensity peak 3 distance 3;; I3D2, normalized light backscatter intensity peak 3 distance 2, k1D1, wavelength peak 1 distance 3; k1D2, wavelength peak 1 distance 2; k2D3, wavelength peak 2 distance 3; b0, b1, b2, b3, b4: regression coefficients; D1,2,3: distances 1 (2 mm), 2 (2.5 mm) and 3 (3 mm); L*, Lightness; b*: Yellowness. Hab: Hue angle; * P < 0.05. ** P < 0.01. *** P < 0.001.
3.3. Prediction of texture Tables 6–9 show the predictive models for: hardness, cohesiveness, damageability and breaking force, respectively. All the texture parameters tested were predicted by the exclusive use of optical parameters and CIELAB coordinates. As shown in Table 6, the best single variable model (model I) contained the independent variable L* and an intercept. This model explained 55% of the variability observed in hardness. The inclusion of Hab, I3D3, k1D1, k2D3, I3D2 in Model I increased the coefficient
of determination (R2), from 0.55 to 0.97 (Model V). According to these results, hue, normalized intensity and wavelength, show potential for online prediction of hardness. As CL prediction, our predictions only used optical parameters and there is no need for including compositional parameters of experimental conditions as predictors. The result in Tables 7 and 8 for cohesiveness and damageability are in the same line that hardness. In both cases only optical parameters and color coordinates were used in the prediction models. Model V for cohesiveness had an R2 of 0.89 while Model 5 for damageability had an R2 of 0.84.
137
G. Nieto et al. / Journal of Food Engineering 143 (2014) 132–138 Table 7 Models for the prediction of cohesiveness in meat emulsions. Model ***
I II*** III*** IV*** V***
b0 C = b0 + k1D2b1 C = b0 + I3D1b1 + I3D3b2 C = b0 + I3D1b1 + I3D2b2 + k1D2b3 C = b0 + I3D1b1 + I3D2b2 + k1D2b3 + I2D3b4 C = b0 + I3D1b1 + k1D1b2 + Habb3 + I2D3b4 + I3D2b5
b1 ns
2.59 0.87*** 2.28ns 2.61ns 3.84**
*
0.07 0.001* 0.001* 0.001* 0.001**
b2
b3
b4
b5
R2
– 0.001* 0.001* 0.001ns 0.01**
– – 0.01ns 0.01* 0.002*
– – – 0.001ns 0.001*
– – – – 0.004***
0.36 0.54 0.70 0.75 0.89
k1D2, wavelength peak 1 distance 2; k1D2, wavelength peak 1 distance 2; I3D1; normalized light backscatter intensity peak 3 distance 1; I3D3, normalized light backscatter intensity peak 3 distance 3; I3D2, normalized light backscatter intensity peak 3 distance 2; I2D3, normalized light backscatter intensity peak 2 distance 3; b0, b1, b2, b3, b4: regression coefficients; D1,2,3: distances 1 (2 mm), 2 (2.5 mm) and 3 (3 mm); Hab: Hue angle; * P < 0.05. ** P < 0.01. *** P < 0.001.
Table 8 Models for the prediction of damageability in meat emulsions. Model ***
I II*** III*** IV*** V***
b0 D = b0 + k1D2b1 D = b0 + I3D1b1 + I3D2b2 D = b0 + I3D1b1 + I3D2b2 + I1D1b3 D = b0 + I3D1b1 + I3D2b2 + k3D3b3 + Cabb4 D = b0 + I3D1b1 + I3D2b2 + k3D3b3 + Cabb4 + I2D3b5
b1 *
179 6.58*** 9.43*** 15.0ns 15.9ns
ns
0.34 0.04** 0.05** 0.07** 0.07**
b2
b3
b4
b5
R2
– 0.06** 0.06** 0.11** 0.11**
– – 0.04ns 0.03ns 0.03*
– – – 0.35* 0.35*
– – – – 0.03ns
0.32 0.56 0.97 0.80 0.84
k1D2, wavelength peak 1 distance 2; k3D3, wavelength peak 3 distance 3; I3D1; normalized light backscatter intensity peak 3 distance 1; I3D2, normalized light backscatter intensity peak 3 distance 2; I3D3, normalized light backscatter intensity peak 3 distance 3; b0, b1, b2, b3, b4: regression coefficients; D1,2,3: distances 1 (2 mm), 2 (2.5 mm) and 3 (3 mm); L*, Cab: Chroma; * P < 0.05. ** P < 0.01. *** P < 0.001.
Table 9 Models for the prediction of breaking force in meat emulsions. Model I*** II*** III*** IV*** V***
BF = b0 + I3D3b1 BF = b0 + k1D2b1 + I3D2b2 BF = b0 + k1D2b1 + I3D1b2 + I3D2b3 BF = b0 + k1D2b1 + k2D1b2 + I3D1b3 + I3D2b4 BF = b0 + k1D2b1 + k1D3b2 + k2D1b3 + I3D1b4 + I3
D2b5
b0
b1
b2
b3
b4
b5
R2
4.93* 449** 330** 1049* 1190**
0.04*** 0.87** 0.64** 1.34*** 1.77***
– 0.03*** 3.15** 3.10** 0.35ns
– – 0.05** 0.05*** 3.40 **
– – –
– – – – 0.10***
0.81 0.90 0.96 0.98 0.99
0.11*** 0.05***
I3D3, normalized light backscatter intensity peak 3 distance 3; I3D2, normalized light backscatter intensity peak 3 distance 2; I3D1, normalized light backscatter intensity peak 3 distance 1; k1D2, wavelength peak 1 distance 2; k1D3; wavelength peak 1 distance 3; b0, b1, b2, b3, b4: regression coefficients; D1,2,3: distances 1 (2 mm), 2 (2.5 mm) and 3 (3 mm); * P < 0.05. ** P < 0.01. *** P < 0.001.
An excellent prediction was obtained for breaking force using 5 optical parameters and an intercept (Model V; R2 = 0.99; Table 9). However, as it can be observed in Table 9, Model 1 using only one predictor (I3) and an intercept had also a high R2 value, 0.81. Taking into account the results, the models explain 97% (Hardness: model V shown in Fig. 2c), 89% (Cohesiveness: model V shown in Fig. 2d), 97% (Damageability: model III shown in Fig. 2e) and 99% (Breaking force: model V shown in Fig. 2f) of the variability observed for the corresponding parameter. Our results are similar to those obtained by Allais et al. (2004) who suggested that prediction found a correlation between final texture parameters of meat emulsions and fluorescence spectroscopy measurements. However, these authors did not present any predictive equation. It is important to note the importance of proper texture for the commercialization of the product. Textural characteristics of food are important aspects of consumer acceptance. Based on consumers demand and the requirements of health authorities, the food research and the meat industry have developed new formulations by reducing or replacing the fat content (Mendoza et al., 2001). However, fermented sausages are among the meat products where
fat reduction is more complicated because fat confers important properties i.e. flavor, and texture (Mendoza et al., 2001). One of the most detrimental effects of fat reduction in fermented sausages is sensory texture (Muguerza et al., 2002). Therefore, the modification of the composition of fermented sausages by reducing its fat content can lead to negative changes in the sensory characteristics of these products and affect consumers’ acceptability. These results show that the use of optical sensor for the study of the light backscatter extinction in finely comminuted meat emulsions may have potential for predicting the texture of the final product. The proposed models help to optimize the texture of the meat emulsion in order to improve the final product.
4. Conclusions In this study, a new on-line optical sensor technology to select the optimum end-point of chopping that would be able to minimize cooking losses and maximize the yield was designed. It could be concluded that the use of a fiber optic light scatter sensor to monitor fresh meat emulsification during the chopping process
138
G. Nieto et al. / Journal of Food Engineering 143 (2014) 132–138
allows successfully predict, cooking losses, textural parameters and lipid oxidation tendency in the cooked product using optical parameters only. The proposed optical measurement system and obtained predictive equations strongly suggest that the development of an inline light scatter sensor for chopping time optimization is feasible. Further research and development efforts are needed to improve accuracy of the predictions and develop a dedicated inline fiber optic sensor. Acknowledgements The authors wish to thank the University of Murcia and the Department of Biosystems and Agricultural Engineering (University of Kentucky) for the economical support of this research and the Departments of Animal and Food Sciences and Biosystems and Agricultural Engineering (University of Kentucky) for allowing the use of their facilities. References Allais, I., Christophe, V., Pierre, A., Dufour, E., 2004. A rapid method based on frontface flurescence spectroscopy for the monitoring of the texture of meat emulsions and frankfurters. Meat Sci. 67, 219–229. Álvarez, D., Payne, F.A., Castillo, M., Xiong, Y.L., 2007. Application of backscatter Light extintion to determine the stability of beef emulsions with different fat/ lean ratios. ASABE, Annual International Meeting, Minneapolis, Minnesota, USA. Álvarez, D., Castillo, M., Payne, F.A., Xiong, Y.L., 2009. A novel fiber optic sensor to monitor beef emulsion stability using visible Light scattering. Meat Sci. 81, 456– 466. Álvarez, D., Castillo, M., Payne, F.A., Cox, Y.L., Xiong, Y.L., 2010. Application of light extinction to determine stability of beef emulsions. J. Food Eng. 96, 309–315.
Barbut, S., 1998. Use of a fiber optic probe to predict meat emulsion breakdown. Ital. J. Food Sci. 3, 253–259. Bendall, J.R., Swatland, H.J., 1988. A review of the relationships of pH with physical aspects of pork quality. Meat Sci. 24, 85–126. Bourne, M.C., 1978. Texture profile analysis. Food Technol. 41 (2), 163–178. Castillo, M., Payne, F.A., Shea, A., 2005. Development of a combined sensor technology for monitoring coagulation and syneresis operations in cheese making. In: Proceedings of ADSA annual meeting. Cincinnati, Ohio, USA. Hunt, R.W.G., 1977. The specification of colour appearance: I. Concepts and terms. Color Res. Appl. 2, 55–68. Jones, K.W., 1984. Protein-lipid interactions in processed meats. In: Proceedings Recipe Meat Conference, vol. 37, p. 52. Koolmes, P.A., Winjgaards, G., Tersteeg, M.H.G., Van Logtestijn, J.G., 1993. Changes in the microstructure of a comminuted meat system during heating. Food Struct. 12, 427–441. Mendoza, E., García, M.L., Casas, C., Selgas, M.D., 2001. Inulin as fat substitute in low fat, dry fermented sausages. Meat Sci. 57, 387–393. Muguerza, E., Fista, G., Ansorena, D., Astiasaran, I., Bloukas, J.G., 2002. Effect of fat level and partial replacement of pork backfat with olive oil on processing and quality characteristics of fermented sausages. Meat Sci. 61, 397–404. NHDSC, 2008. National Hot Dog and Sausage Council. Vital Hot Dogs Statistics. AMI Foundation, Washington, USA. Nieto, G., Castillo, M., Xiong, Y.L., Alvarez, D., Payne, F., Garrido, M.D., 2009. Antioxidants and emulsifying properties of alcalase-hydrolyzed potato proteins in meat emulsions with different fat concentration. Meat Sci. 83 (1), 24–30. SAS, 2002. Statistical Analysis System, Rel. 9.1. SAS Institute Inc., Cary, NC. Swatland, H.J., 2002. On-line monitoring of meat quality. In: Kerry, J., Kerry, J., Ledward, D. (Eds.), Meat Processing Improving Quality. Woodhead Publishing, pp. 193–216. USDA (United States Department of Agriculture), 2005. Meat Animal’s Production Disposition and Income 2004 Summary. Wang, L.L., Xiong, Y.L., 2005. Inhibition of lipid oxidation in cooked beef patties by hydrolyzed potato protein is related to its reducing and radical scavenging ability. J. Agricult. Food Chem. 53, 9186–9192. Xiong, Y.L., Noel, D.C., Moody, W.G., 1999. Textural and sensory properties of lowfat beef sausages with added water and polysaccharides as affected by pH and salt. J. Food Sci. 64, 550–554.