LWT - Food Science and Technology 137 (2021) 110562 Contents lists available at ScienceDirect LWT journal homepage: www.elsevier.com/locate/lwt An in-line optical technology to control the emulsification degree in a continuous industrial emulsifier for meat sausage production Zulay González-Martínez, Anna Zamora *, Montserrat Mor-Mur, Manuel Castillo ** Centre d’Innovació, Recerca i Transferència en Tecnologia dels Aliments (CIRTTA), Department of Animal and Food Sciences, Universitat Autònoma de Barcelona, 08193, Bellaterra, (Cerdanyola del Vallès), Barcelona, Spain A R T I C L E I N F O A B S T R A C T Keywords: The stability of comminuted products plays an important role in the economy of meat industries. Proper Meat emulsion formulation and the establishment of suitable emulsification conditions can significantly contribute to cooking Cooking loss losses control. The purpose of this research was to study the relationship between light backscatter parameters Emulsification degree and cooking losses and develop prediction models that would allow the optimization of the emulsification Light backscatter Prediction models process in a continuous industrial emulsifier. The optical response of meat emulsions, produced at industrial scale, depended drastically on the formulation (with and without starch) and the degree of emulsification. Furthermore, formula with starch showed significantly lower cooking losses than formula without starch, but for both formulas, several optical parameters correlated with cooking losses. Models for the prediction of cooking losses with R2 values > 0.999 were obtained with five or six statistically significant optical predictors depending on the formula. These results point out the potential of light backscatter technology as a control tool during emulsification. 1. Introduction (over-processing), as a result of excessive emulsion temperature (Feiner, 2006, pp. 239–286; Knipe, 2014); and c) product with visible fat par­ Meat emulsions are the result of the comminution of water, proteins, ticles (under-processing), having incomplete solubilized proteins and/or and fat (Puolanne, 2010). During emulsification, myosin solubilized salts. The stability of pre-cooked emulsions defines gel characteristics. protein covers fat particles to prevent fat separation through thermal During cooking, proteins change their conformational structure pro­ treatment, while actin solubilized protein immobilizes water. The moting aggregation. Shrinkage provokes gel deformation and concom­ interaction between proteins, fat, and water set up a tridimensional itant cooking losses (Tornberg, 2005) which have been reported emulsion matrix (Barbut, 1995; Feiner, 2006, pp. 239–286). Composi­ between 5 and 18% in frankfurters (Grigelmo-Miguel, Abadías-Serós, & tion and processing conditions determine emulsions stability which has Martín-Belloso, 1999; Shan et al., 2014), representing 0.38–1.36 a direct effect on water holding capacity or cooking losses (Sebranek, American billion US dollars calculated with the index 0.076 American 2003). In some cases, starch is used to improve water and protein billion US dollars per 1% loss of Álvarez, Castillo, Payne, Cox, and Xiong binding, hence increasing yield and profitability. Meat emulsions with (2010). starch show a compact network where water is better retained due to There are few works on emulsion stability control through optical starch swelling ability and the interaction with meat proteins (García-­ sensor technologies based on light backscatter (Álvarez, Castillo, Payne, García & Totosaus, 2008). & Xiong, 2009, 2007;Álvarez, Castillo, Payne, et al., 2010; Álvarez, Optimizing the emulsification degree requires avoiding three Castillo, Xiong, & Payne, 2010; Nieto, Xiong, Payne, & Castillo, 2015, possible processing-related defects: a) less firm unstable product with 2014; Torres, 2016). The implementation of this type of technology undue fat surface area (over-processing), which enhances water and fat control could favor monitoring during emulsification, to determine in separation (Hoogenkamp, 2011); b) product with fat melting problems time, the homogenization end-point or speed and prevent cooking * Corresponding author. Departament de ciència animal i dels aliments, Facultat de Veterinària, Universitat Autònoma de Barcelona, Travessera dels Turons, Edifici V, 08193 Bellaterra (Cerdanyola del Vallès), Barcelona, Spain. ** Corresponding author. Departament de ciència animal i dels aliments, Facultat de Veterinària, Universitat Autònoma de Barcelona, Travessera dels Turons, Edifici V, 08193 Bellaterra (Cerdanyola del Vallès), Barcelona, Spain. E-mail addresses: anna.zamora@uab.cat (A. Zamora), manuel.castillo@uab.es (M. Castillo). https://doi.org/10.1016/j.lwt.2020.110562 Received 22 May 2020; Received in revised form 9 October 2020; Accepted 7 November 2020 Available online 11 November 2020 0023-6438/© 2020 Elsevier Ltd. This is an open access article under the CC BY-NC-ND license (http://creativecommons.org/licenses/by-nc-nd/4.0/). Z. González-Martínez et al. LWT 137 (2021) 110562 Fig. 1. Typical optical spectrum and the identification of eight basal predictors. losses. In that way, products would not be affected by emulsion break­ continuous industrial emulsifier (Model I175CDVM-90D, INOTEC, down improving yield and quality (Nieto, Xiong, Payne, & Castillo, Reutlingen, Germany), where the emulsification process occurred. 2014). The optical device proposed implements correlations of some Different degrees of emulsification were obtained by means of volu­ color and optic parameters with water or fat losses to determine the metric flow adjustment, and monitored through the temperature of the exact emulsification end-point in meat emulsions. These studies have product during extrusion: under-processed samples (5.09 ± 0.23 ◦ C), demonstrated the relation between cooking losses and the optical samples with medium-degree of processing (hereafter referred as to response. All these studies, except for Torres (2016), were manufactured medium-processed samples; 7.41 ± 0.70 ◦ C), and over-processed sam­ under laboratory/pilot plant processing, so the effect of real industrial ples (9.36 ± 0.48 ◦ C). One pre-emulsified sample and the three samples conditions on the optical technology feasibility has not been analyzed with different emulsification degrees constituted one replicate. Proxi­ yet. mate composition of emulsions with and without starch, characterized In order to achieve faster comminution and processing, conventional with a Food Scan NIR Meat Analyzer (DK-3400, FOSS, Hillerod, cutters are being replaced by continuous emulsifiers with a coarse Denmark), was approximate: 62.82 and 65.92% of moisture, 13.21 and chopping step followed by the emulsification step executed with a 17.29% of fat, 11.57 and 13.13% of proteins and 2.12 and 2.10% of salt, combination of blades and perforated plates (Morin, Reeve, Tomey, respectively. All samples were vacuum packaged and delivered refrig­ Wilke, & Lucke, 2005; Powers, Schack, & Anderson, 1987; Rudibaugh, erated. Analyses were performed on the same day of reception at UAB or 1995). None of the previous studies have used continuous emulsifiers. In the next day (storage at 4 ± 2 ◦ C overnight). that view, the present study had the purpose to provide valuable in­ formation with the use of industrial meat samples thus going forward 2.2. Meat emulsion cooking losses towards a new in-line control system. The present work aimed to find out a relation between some optical parameters (excluding color co­ Cooking losses were measured using a modified method of Bañón, ordinates for practical purposes), and the cooking losses of two different Díaz, Nieto, Castillo, and Álvarez (2008) and Nieto et al. (2009), where meat emulsions (with and without starch) produced through continuous meat emulsions were introduced into a syringe barrel of 100 mL and emulsification, to establish prediction equations for the losses in both pressed inside a weighted 50 mL corning tube with a plunger, to simu­ types of samples, and, as a consequence, to evaluate the feasibility of late casing stuffing. Then, each corning tube was weighted again and applying the optical backscatter technology as a control technology for placed in a water bath (Ovan Inox 27 L Model, Suministros Grupo Esper the emulsification degree in meat emulsions. S.L., Barcelona, Spain) at 75 ± 1 ◦ C for 45 min. After cooking, all tubes were placed inverted on a metal mesh during 1 min to drain the expelled 2. Materials and methods liquid and finally weighted. ( ) 2.1. Meat emulsion manufacture and composition Cooking losses were calculated with the formula CL = W0W− 0Wf ⋅100, where W0 was initial emulsion weight and Wf final cooked emulsion Two types of meat emulsions (with and without 10% potato starch) weight. Each trial was performed sixfold. were produced at industrial scale following standard procedures by a large pioneer company in the meat industry. Lean trimmings (pork, chicken, and turkey), fat (pork fat trimmings and rind), salt, spices, 2.3. Light backscatter measurement of meat emulsions potato starch, and functional additives were purchased from suppliers approved by the European Union. Appropriate proportions, not speci­ The experiment was carried out on a High-Resolution Fiber Optic fied by the meat company due to confidentiality on industrial formula­ Spectrometer (Model HR4000, Ocean Optics, Inc., Dunedin, FL, USA) tions, were calculated to obtain a base theoretic formula of 70, 17, and fed by a tungsten halogen bulb (300–1100 nm) as light source (LS-1, 13% of moisture, fat, and proteins. All required ingredients were mixed Ocean Optics, Inc.) and communicated with a double-jacketed sample using an industrial mixer (Model IM-4500, INOTEC, Reutlingen, Ger­ holder through two fiber optic cables of ~600 μm diameter each (Ocean many) to obtain pre-emulsified batters which were introduced into a Optics, Inc.). Both optic fiber ends were attached to a light backscatter probe and coupled to the sample holder; the other two ends were 2 Z. González-Martínez et al. LWT 137 (2021) 110562 connected to the spectrometer and the light source, respectively. This values of pre-emulsified samples, per production, were subtracted from system delivered optical data from the spectrometer to the Spec­ the respective data of each type of emulsion (under-, medium- and traSuite® software (Ocean Optics, Inc.) to obtain light backscatter over-processed). From the normalized basal predictors, the ratio of spectra. peaks, the ratio of inflection points, and their mathematical trans­ The acquisition time was set to 3 s (i.e., integration time). Each meat formations (inverse, square, and cube) were calculated only with in­ emulsion was stuffed into the sample holder, the light source blocked tensity data. The ratio between the intensity at 450 nm and peaks or and the sensor probe placed until the head of the sensor reached the inflection points were also studied. With this described procedure, a emulsion. After that, the options “scope minus dark”, “store dark” and total of 190 predictors were obtained and distributed in three blocks “store dark minus” were set, consecutively, in order to subtract the named “Peaks & inflection points” (wavelength and intensity values of background noise. Then, the light source was unblocked and the sample peaks and inflection points and their mathematical transformations), spectrum saved when stable. Data from optical spectra were collected at “Ratios of peaks” (ratios of the intensities of peaks and/or the intensity least sixfold per sample. In the range of 420–635 nm, eight predictors at 450 nm, and their mathematical transformations) and “Ratios of in­ were identified as peaks (maxima on the light backscatter intensity flection points” (ratios of the intensities at the inflection points and/or spectra) and inflection points (maximum rate of light backscatter in­ the intensity at 450 nm, and their mathematical transformations). Data tensity increase identified as maxima on the first derivative vs. wave­ was analyzed in blocks in order to evaluate the accuracy of prediction length curve) (Fig. 1). Data was analyzed as intensity and wavelength for obtained with regard to the readiness of predictors calculation. peaks and inflection points. A non-variable zone in the spectra was detected at 450 nm and was used as a normalization factor. To exclude 2.4. Statistical analysis the composition influence in the cooking losses observed in other works (Allais, Viaud, Pierre, & Dufour, 2004; Bañón et al., 2008) average The whole experiment was repeated on four and three independent occasions for formula with and without starch, respectively. Analysis of variance (ANOVA) was performed with the Statistical Analysis System (SAS) in order to investigate the effect of emulsification degree, the Table 1 process factor, and the emulsion production batch on optical predictors Cooking losses (%) depending on emulsification degree and formula. and cooking losses, including into the statistical model both factors and Emulsion Emulsification degree their interaction. LSD test was used for comparison of sample data, and Under- Medium- Over- evaluations were based on a significance level of (P < 0.05). Further­ processed processed processed more, Pearson’s correlation coefficients between optical predictors and Formula with starch1 3.996 ± 1.580a 3.472 ± 1.568b 4.078 ± 1.793a cooking losses were determined. Different regression models for pre­ Formula without 4.884 ± 1.711a 5.471 ± 2.363a 4.774 ± 1.151a dicting cooking losses with the optical predictors were tested using the starch2 maximum R2 procedure of SAS to obtain the best eight models of Mean value ± s.d.; 1n = 72; 2n = 54; a, b: values by rows with different super­ cooking losses prediction for each of the three different blocks of data as script letter were significantly different (P < 0.05). well as for all data together. Table 2 Models for the prediction of cooking losses in meat emulsions with starch (ns, not significant P ≥ 0.05). Data Model Equation R2 Peaks & inflection points I** Closs = 3.60 + 1.62P7 0.623 II*** Closs = 4.38 + 1.36P7 - 2.44P82 0.902 III*** Closs = 4.98 + 1.11P7 - 1.81P82 - 0.462P2 0.964 IV*** Closs = 5.82 + 0.888P7 - 0.879P2 - 2.18⋅10− 6P83 - 1.02P8 0.989 V*** Closs = 5.66 + 0.665P7 - 0.870P2 - 0.754P8 - 9.10⋅10− 9P103 + 0.157P6 0.996 VI*** Closs = 7.28–1.84P2 - 6⋅10− 8P103 - 0.00179P13 - 0.0880P80 - 0.109P81 + 4.38⋅10− 6P84 >0.999 VII*** Closs = 6.90–1.75P2 - 6.07⋅10− 8P103 + 0.0859P6 - 0.0684P80 - 0.121P81 + 3.98⋅10− 6P84 - 3.47P68 >0.999 VIII*** Closs = 6.85–1.74P2 - 6.03⋅10− 8P103 + 0.0890P6 - 0.0678P80 - 0.121P81 + 3.96⋅10− 6P84 - 3.57P68 + 0.0226Pns 61 >0.999 Ratios of peaks I** Closs = 5.08–25.3P18 0.556 II*** Closs = 6.25–55.2P18 + 4460.7P123 0.820 III*** Closs = 5.51–185.6P18 - 153.8P17 + 254.7P108 0.900 IV*** Closs = 5.39–207.7P18 - 178.8P17 + 265.2P108 + 2.01Pns126 0.913 V*** Closs = 6.84–170.8P18 - 90.6Pns 17 - 16.9P22 + 138.2P122 + 288.1P107 0.968 VI*** Closs = 7.87–100.8P18 + 486.7P122 + 718.8P107 - 2.30P115 - 718,849P129 + 80P19 0.993 VII*** Closs = 7.78–100P18 + 564.8P122 + 672.5P107 - 2.34P115 - 904,374P129 + 79.2P19 - 0.184Pns 127 0.996 VIII*** Closs = 7.49–70.9P18 + 15.4P126 + 530.7P122 - 0.113P115 - 716,421P129 + 74P19 - 0.256P127 + 44.5P112 >0.999 Ratios of inflection points I** Closs = 5.80 + 56.9P45 0.755 II*** Closs = 6.50 + 62.2P45 - 40.6P142 0.936 III*** Closs = 6.35 + 63.9P45 - 32.8P142 + 467.2P180 0.964 IV*** Closs = 6.37 + 80.7P45 - 30.1P142 + 892.1P180 + 100.3Pns 145 0.977 V*** Closs = 6.62 + 95.6P45 - 46.1P142 + 167P145 + 8983.3P171 + 9.88P184 0.989 VI*** Closs = 6.55 + 104.7P45 - 46.8P142 + 336.3P145 + 11,116P171 + 11.5P184 - 167.2Pns 176 0.993 VII*** Closs = 6.46 + 98.9P45 - 42.7P142 + 352.8P145 + 13,758P171 + 7.20P184 - 248P176 + 2.64P42 0.998 VIII*** Closs = 6.69 + 96.6P45 - 46.2P142 + 477P145 + 13,334P171 + 13.5P184 - 325.5P176 - 11.0P37 + 38.5P57 >0.999 All data I** Closs = 5.80 + 56.9P45 0.755 II*** Closs = 6.50 + 62.2P45 - 40.6P142 0.936 III*** Closs = 5.59–31.6P142 - 8.54P46 + 0.673P7 0.983 4 IV*** Closs = 5.84–32.9P142 - 8.33P46 + 0.589P7 - 5.39⋅10− P10 0.997 4 V*** Closs = 5.86–34.9P142 - 8.21P46 + 0.613P7 - 5.47⋅10− P10 - 12.1P174 0.999 4 VI*** Closs = 6.33–43.2P142 - 8.71P46 + 0.671P7 - 5.58⋅10− P10 - 21.9P174 - 0.270P61 >0.999 4 VII*** Closs = 6.26–43.3P142 - 7.57P46 + 0.677P7 - 5.53⋅10− P10 - 22.4P174 - 0.260P61 - 79.3P147 >0.999 4 VIII*** Closs = 6.21–42.4P142 - 6.64P46 + 0.656P7 - 5.55⋅10− P10 - 19.8P174 - 0.233P61 - 237.9P147 - 1257.4P177 >0.999 3 Z. González-Martínez et al. LWT 137 (2021) 110562 3. Results and discussion differentiated by some specific predictors. Although none of the pre­ dictors could differentiate the three emulsification degrees, a large 3.1. Meat emulsions composition and cooking losses number could identify and set apart one emulsification degree as different from the other two (data not shown). Almost all of these pre­ Starch addition could elucidate cooking losses differences observed dictors were found in the “Peaks & inflection points” block, a fact of between emulsions with and without starch (Table 1). In fact, Chen, Lee, interest as the majority of predictors that correlated with cooking losses and Crapo (1993) showed that starch embedded in a protein gel matrix were found in this block (predictors 2, 6, 80). As a result, the above- swelled during cooking and enhanced the formation of strong structures, mentioned predictors were part of the cooking losses prediction which is represented by a more stable matrix with greater water-binding equations. capacity. Other studies have also evidenced the ability of starch to It was observed that the Pearson correlation values of some of the entrap water and the consequent reduction in cooking losses when it is predictors did not correlate significantly (P ≥ 0.05) with the losses (data incorporated like in low-fat bologna sausages, low-fat frankfurters, and not shown). This fact could be explained by their low contribution, i.e. other pork batters (Bañón et al., 2008; Dexter, Sofos, & Schmidt, 1993; little information, but when included in the model they potentiated the Hughes, Mullen, & Troy, 1998). It should be considered that in the results since the R2 values increased significantly. In this way, for the previous studies, meat emulsions were processed by bowl choppers, “Peaks & inflection points” block, the predictors 7, 82, 2, 83, and 8 which differs from the equipment used in this study. Independently of explained the first four models, with R2 of 0.623, 0.902, 0.964, and the process, the same tendency of losses was found in the present study 0.989, respectively (Table 2). It can also be noticed that the inclusion of for the case of meat emulsions with starch (Table 1). just two variables (Model II***) improved notably the determination A study in meat emulsions and frankfurters pointed out increases in coefficient (R2 = 0.902) when compared to Model I** (R2 = 0.623), the losses when the chopping time passed from 3 to 7 min at 2000 and which suggested that only two predictors could be enough to have a 3000 rpm in a 30 L-Stephan pilot equipment, respectively (Allais et al., representative cooking loss prediction. Similar results were reported in a 2004). Similarly, other studies have borne out that an increment in the study made on fresh pork meat emulsions formulated with hydrolyzed chopping time increased cooking losses in comminuted pork meats potato protein and different fat levels (15% and 30%) with coefficients (Bañón et al., 2008; Álvarez et al., 2007). These findings showed the of determination of 0.77, 0.95, and 0.96 in the first three models when importance of chopping as a factor in the control of meat emulsions color parameters (L* and b*) and the optic parameter peak2 wavelength stability. However, in the present study, no clear tendency was observed were incorporated in the models (Nieto et al., 2009, 2014). It should be when analyzing the effect of the emulsification degree on cooking losses. noted, though, this model did include color parameters, which are less For example, in samples with starch, a significant difference was convenient for the point of view of sensor technology as compared to observed between the medium-degree of processing and the other two static light scatter readings. degrees, which is also where the lowest cooking loss occurred (3.47%). All predictors for blocks “Ratios of peaks” and “Ratios of inflection This difference can be attributed to the know-how of the industry, which points” were included in their respective cooking losses prediction has established medium-processed conditions in starch sausages as the models. On one hand, for the case of “Ratios of peaks”, predictors 108 optima in quality and cost-effective results. On the contrary, there were and 126 were the only ones that correlated significantly with cooking no significant differences between emulsification degrees for cooking losses (P < 0.05) (Table 2 Supplementary data), but these showed up losses in emulsions without starch, maybe due to erratic standard de­ only in Models III*** (R2 = 0.900), IV*** (R2 = 0.913) and VIII*** (R2 > viations observed in this type of emulsion, which could interfere in the 0.999) (Table 2). On the other hand, for the case of “Ratios of inflection visualization of the differences (Table 1). points”, there were no predictors that correlated with losses, however, the models showed high determination coefficients starting from Model II*** with R2 of 0.936 (Table 2). This suggests that some predictors by 3.2. Optical predictors, correlations, and cooking loss prediction themselves contribute with little information but when included in the equations models cause a notable improvement in the R2 value. Finally, the potential estimation of predictors from the three blocks 3.2.1. Emulsions with starch together was analyzed. The cooking losses prediction models showed R2 The ANOVA analysis performed with the whole set of optical data values of 0.997 for Model IV***, 0.999 for Model V*** and >0.999 for showed that the emulsification degree could be statistically Fig. 2. A) Predicted values of cooking losses for meat emulsions with starch by Model VI (P ≤ 0.001) with “Peaks & inflection points” block. B) Predicted values of cooking losses for meat emulsions without starch by Model V (P ≤ 0.001) with “All data” block. R2: determination coefficient; SEP: standard error of prediction (%); CV: coefficient of variation (%). 4 Z. González-Martínez et al. LWT 137 (2021) 110562 Models VI***, VII*** and VIII*** (Table 2). information. Indeed, predictors 94, 105, and 100 of “Peaks & inflection These results suggest that, regarding meat emulsions with starch, the points” block were included within the Models I**, II** and III** when optical information of “Ratios of inflection points” and their trans­ all the blocks were analyzed together (Table 3). formations was valuable when establishing prediction equations with a Particularly, some of the predictors did not show a significant cor­ small number of predictors (from 1 to 3; Table 2). However, using only relation value with cooking losses, which could be attributed to their the data from this block, 8 predictors were needed to reach a determi­ little information by themselves; but when included in the models the nation coefficient >0.999. In fact, the best model to reach a determi­ coefficients of determination (R2) improved significantly. nation coefficient >0.999 with less number of variables was Model In addition, it can be noted that Model V*** of “All data” block VI*** with 6 predictors of the “Peaks & inflection points” block, which is (Table 3) reached the maximum determination coefficient (R2 > 0.999) represented in Fig. 2A. using 5 predictors which suggests that 5 optical predictors could be enough to represent, virtually without error, the cooking loss in meat 3.2.2. Emulsions without starch emulsions without starch. The representation of predicted vs. experi­ Concerning emulsions without starch, it was found that at least one mental cooking losses using Model V*** is shown in Fig. 2B. type of emulsification degree could be differentiated from the rest by some of the predictors (data not shown). This trend was widely found in 3.2.3. With starch vs. without starch emulsions the three blocks; however, at the “Peaks & inflection points” block, only As already mentioned, emulsions with and without starch were four predictors (11, 69, 85, and 101) were able to differentiate indi­ characterized by different types of predictors, which notably tended to vidually the three emulsification degrees. These four predictors came be less informative in the emulsions with starch when the effect of from the same optical parameter and the corresponding trans­ emulsification degree on the optical parameters and their correlations formations. Furthermore, looking at Pearson’s coefficients, although with cooking losses was studied. The contrary was found in samples these predictors did not correlate with cooking losses, many other pre­ without starch were a wide variety of predictors, including one group of dictors in all the three blocks showed significant correlations with predictors that differentiated the three emulsification degrees, provided cooking losses (predictors 100, 105, 19, 108, 110, 28, 40, 47, 173, 150). strong information about cooking losses and the emulsification degree. The following cooking losses prediction models and the corre­ The reduced number of predictors correlating with the losses in emul­ sponding regression coefficients for “Peaks & inflection points”, “Ratios sions with starch could be a consequence of starch incorporation, given of peaks” and “Ratios of inflection points” (Table 3) showed that for a that it improves notably the stability of the matrix emulsion by pro­ three-variable model the regression coefficients were 0.917 with the moting the interaction between the main components of the batter predictors 105, 100, 4 for the “Peaks & inflection points”, 0.900 with the (Dexter et al., 1993). Probably, this made the emulsions more homog­ predictors 19, 110, 119 for “Ratios of peaks” and 0.918 with the pre­ enous providing similar and reliable optical data during the light dictors 156, 161, 47 for “Ratios of inflection points”, respectively. backscatter scanning and overshadowing some strong predictors that in The highest determination coefficients were shown in the “Peaks & the models seemed to be significant to predict the losses. The opposed inflection points” and “Ratios of inflection points” blocks. Nevertheless, situation may have occurred in emulsions without starch, where a more the fact that Pearson’s coefficient of some of the “Ratios of inflection heterogeneous matrix may have been obtained (Lyons, Kerry, Morrissey, points” predictors mentioned above were not significant suggests that & Buckley, 1999). So, the effect of the emulsification degree clearly “Peaks & inflection points” prediction models had more valuable found in emulsions without starch may be suppressed when adding it. Table 3 Models for the prediction of cooking losses in meat emulsions without starch (ns, not significant P ≥ 0.05). Data Model Equation R2 Peaks & inflection points I** Closs = 5.48–0.276P94 0.658 II** Closs = 6.50–0.430P94 - 7.12⋅10− 11Pns 105 0.806 III** Closs = 5.73–1.03⋅10− 9P105 + 7.32⋅10− 10P100 - 1.32P4 0.917 IV** Closs = 5.91–1.22⋅10 P105 + 8.78⋅10− 10P100 - 1.45P4 - 0.0649Pns − 9 75 0.970 V*** Closs = 6.32–1.26⋅10− 9P105 + 9.28⋅10− 10P100 - 1.87P4 - 0.0735P75 - 4.93⋅10− 3P12 0.997 VI*** Closs = 6.11–1.40⋅10 P105 + 1.035⋅10 P100 - 1.83P4 - 0.0713P75 - 5.22⋅10− 3P12 + 0.138P64 − 9 − 9 >0.999 VII*** Closs = 6.09–1.36⋅10− 9P105 + 1.01⋅10− 9P100 - 1.83P4 - 0.0656P75 - 6.25⋅10− 3P12 + 0.118P64 - 0.0648Pns 62 >0.999 Ratios of peaks I* Closs = 6.13 + 14.066P21 0.556 II* Closs = 6.59 + 50.84P19 + 31.4Pns 110 0.820 III* Closs = 9.49 + 175.5Pns ns ns 19 + 131.9P110 + 4326.1P119 0.900 IV* Closs = 8.45 + 210.6P110 - 1911.6P107 + 4568.4P123 + 266.3P28 0.913 V* Closs = 8.59 + 236.7P110 - 1902.7P107 + 4858.7P123 + 313.6P28 - 303Pns 120 0.968 VI* Closs = 4.06–3,097P107 + 11,199P123 - 1129.1Pns ns 120 + 619P108 + 4764.6P109 - 14,687P118 0.993 VII* Closs = 3.83–8518.9Pns 119 - 3657.5P107 + 11,587P123 - 2283.7P108 + 734.8P108 + 5016.2P109 - 15,588P118 0.996 Ratios of inflection points I* Closs = 5.91–81.5P150 0.605 II** Closs = 5.00 + 128.6P156 + 578.1P161 0.871 III** Closs = 4.27 + 175P156 + 1320.4P161 - 29.5Pns 47 0.918 IV*** Closs = 4.70–657.5P145 - 1,558P147 + 129.6P148 + 755.1P151 0.996 V*** Closs = 4.79–631.4P145 - 1491.6P147 + 122.5P148 + 772.6P151 - 816.8P173 0.999 VI*** Closs = 5.06–653.1P145 - 1457.3P147 + 117.9P148 + 908.8P151 + 3428.1P181 + 0.395P40 >0.999 VII*** Closs = 5.05–653.7P145 - 1463.9P147 + 118.5P148 + 908.5P151 + 3664.5P181 + 0.389P40 + 0.186P54 >0.999 All data I** Closs = 5.48–0.276P94 0.658 II** Closs = 6.50–0.430P94 - 7.12⋅10− 11Pns 105 0.806 III** Closs = 2.87–1.44⋅10− 9P105 + 1.16⋅10− 9P100 + 237.9P4 0.924 IV*** Closs = 4.41–3.26⋅10− 9P105 + 2.49⋅10− 9P100 + 2.85P7 + 33.1P31 0.997 V*** Closs = 4.34–3.25⋅10− 9P105 + 2.49⋅10− 9P100 + 2.75P7 + 32.2P31 - 4.56⋅10− 3P91 >0.999 VI*** Closs = 4.29–3.19⋅10− 9P105 + 2.44⋅10− 9P100 + 2.74P7 + 36.2P31 - 2.62⋅10− 3P91 + 0.0782P6 >0.999 VII*** Closs = 4.29–3.18⋅10− 9P105 + 2.44⋅10− 9P100 + 2.73P7 + 36.73P31 - 2.52⋅10− 3P91 + 0.0802P6 + 3.24⋅10− 3P162 >0.999 5 Z. González-Martínez et al. LWT 137 (2021) 110562 Similar results were found in a previous investigation done in meat Funding sources emulsions with and without starch and light backscatter technology, reporting more predictors in the samples without starch which differ­ This work was supported by the Project COMRDI15-1-027 (Comu­ entiate emulsification degree (Torres, 2016). Related findings were nitat RIS3CAT d’Alimentació INNOÀPAT), co-funded by ACCIÓ suggested in Álvarez et al. (2007) with pork emulsions manufactured at (Catalan Government) and the European Regional Development Fund a laboratory scale with and without starch and at different lean/fat ra­ (ERDF/FEDER). tios. The results showed a clearer response of the studied variable (lightness) concerning the chopping time and cooking losses in emul­ Declaration of competing interest sions without starch. These outcomes together with those of the present study, suggest a better optical response for emulsions without starch None. when different emulsification degrees are applied. However, it is clear from the results, that cooking losses can be optically predicted with a Appendix A. Supplementary data similar accuracy irrespectively of the presence/absence of starch in the formula. Supplementary data to this article can be found online at https://doi. Among all the models for each type of emulsions, the best cooking org/10.1016/j.lwt.2020.110562. losses prediction equations were found in the “Peaks & inflection points” block for formula with starch and in “All data” for formula without References starch (Tables 2–3). These models reached the maximum determination coefficient (R2 > 0.999) with 5 and 6 predictors for the formula without Allais, I., Viaud, C., Pierre, A., & Dufour, É. (2004). A rapid method based on front-face fluorescence spectroscopy for the monitoring of the texture of meat emulsions and and with starch, respectively. The results showed a noticeable frankfurters. Meat Science, 67(2), 219–229. https://doi.org/10.1016/j. improvement in the determination coefficients models proposed by meatsci.2003.10.009 Álvarez et al. (2007) (R2 = 0.69, four predictors model) and Nieto et al. Álvarez, D., Castillo, M., Payne, F. A., Cox, R. B., & Xiong, Y. L. (2010). Application of light extinction to determine stability of beef emulsions. Journal of Food Engineering, (2014) (R2 = 0.97, five predictors model). Although none of the afore­ 96(2), 309–315. https://doi.org/10.1016/j.jfoodeng.2009.08.003 mentioned works matched exactly with the present study conditions, it Álvarez, D., Castillo, M., Payne, F. A., Garrido, M. D., Bañón, S., & Xiong, Y. L. (2007). should be noted that, in the present study, samples were generated using Prediction of meat emulsion stability using reflection photometry. Journal of Food Engineering, 82(3), 310–315. https://doi.org/10.1016/j.jfoodeng.2007.02.031 a continuous industrial emulsifier and no color parameters were used as Álvarez, D., Castillo, M., Payne, F. A., & Xiong, Y. L. (2009). A novel fiber optic sensor to predictors. monitor beef meat emulsion stability using visible light scattering. Meat Science, 81 For the case of the models proposed by Álvarez et al. (2007), the (3), 456–466. https://doi.org/10.1016/j.meatsci.2008.09.007 determination coefficients ranged from 0.42 to 0.69 when two different Álvarez, D., Castillo, M., Xiong, Y. L., & Payne, F. A. (2010). Prediction of beef meat emulsion quality with apparent light backscatter extinction. Food Research types of meat emulsions (starch and no starch) produced at laboratory International, 43(5), 1260–1266. https://doi.org/10.1016/j.foodres.2010.03.002 scale were analyzed. The low R2 found gave sight that the predictors Bañón, S., Díaz, P., Nieto, G., Castillo, M., & Álvarez, D. (2008). Modelling the yield and proposed in their models (chopping time, temperature, and color co­ texture of comminuted pork products using color and temperature. Effect of fat/lean ratio and starch. Meat Science, 80(3), 649–655. https://doi.org/10.1016/j. ordinates) were not sensible enough to predict the cooking losses. meatsci.2008.03.001 Later on, Nieto et al. (2014) incorporated for the first time optical Barbut, S. (1995). Importance of fat emulsification and protein matrix characteristics in spectra parameters in cooking losses prediction models to describe the meat batter stability. Journal of Muscle Foods, 6, 161–177. Chen, J. S., Lee, C. M., & Crapo, C. (1993). Linear programming and response surface optimum end-point of emulsification. Meat samples, manufactured at a methodology to optimize surimi gel texture. Journal of Food Science, 58(3), 535–536. laboratory scale, were formulated with hydrolyzed potato protein and https://doi.org/10.1111/j.1365-2621.1993.tb04318.x analyzed by light backscatter technology. Their results showed an R2 of Dexter, D. R., Sofos, J. N., & Schmidt, G. R. (1993). Quality characteristics of Turkey bologna formulated with carrageenan, starch, milk and soy protein. Journal of Muscle 0.97, much lower than the maximum coefficients of determination (R2 Foods, 4(3), 207–223. https://doi.org/10.1111/j.1745-4573.1993.tb00503.x > 0.999) found in the present study. The authors suggested that the dark Feiner, G. (2006). Cooked sausages. Meat products handbook. Boca Raton: Practical Science color of the hydrolyzed potato protein may have interfered in the optical and Technology. https://doi.org/10.1533/9781845691721.2.239 García-García, E., & Totosaus, A. (2008). Low-fat sodium-reduced sausages: Effect of the response of the emulsion. In the present work, such difficulties were not interaction between locust bean gum, potato starch and carrageenan by a mixture found. design approach. Meat Science, 78(4), 406–413. https://doi.org/10.1016/j. meatsci.2007.07.003 4. Conclusions Grigelmo-Miguel, N., Abadías-Serós, M. I., & Martín-Belloso, O. (1999). Characterisation of low-fat high-dietary fibre frankfurters. Meat Science, 52(3), 247–256. https://doi. org/10.1016/S0309-1740(98)00173-9 The study of the cooking losses and the optical response of two Hoogenkamp, H. W. (2011). Protein performance in emulsion stability. Fleischwirtschaft different industrial meat emulsions allowed the identification of some International, 3, 54–59. Hughes, E., Mullen, A. M., & Troy, D. J. (1998). Effects of fat level, tapioca starch and optical parameters as potential predictors of the cooking losses. This led whey protein on Frankfurters formulated with 5% and 12% fat. Meat Science, 48 to the development of statistically significant prediction equations for (1–2), 169–180. https://doi.org/10.1016/S0309-1740(97)00087-9 the cooking losses with coefficients of determination, R2 > 0.999, in Knipe, C. L. (2014). Types of sausages: Emulsion. In Encyclopedia of meat sciences (pp. 256–260). Elsevier. https://doi.org/10.1016/B978-0-12-384731-7.00143-4. both types of emulsions. These results point out the potential of light Lyons, P. H., Kerry, J. F., Morrissey, P. A., & Buckley, D. J. (1999). The influence of added backscatter technology as a tool to predict cooking losses and suggest whey protein/carrageenan gels and tapioca starch on the textural properties of low the implementation of an in-line/on-line optical emulsification control fat pork sausages. Meat Science, 51(1), 43–52. https://doi.org/10.1016/S0309-1740 (98)00095-3 technology that would significantly contribute to the selection of an Morin, P. G., Reeve, M. L., Tomey, J. L., Wilke, D. B., & Lucke, D. E. (2005). Integrated optimum emulsification degree. continuous meat processing system. In United States patent application publication 11/ 061,717. https://doi.org/10.1111/j.1559-3584.1927.tb04229.x Nieto, G., Castillo, M., Xiong, Y. L., Álvarez, D., Payne, F. A., & Dolores, M. (2009). Antioxidant and emulsifying properties of alcalase-hydrolyzed potato proteins in meat emulsions with different fat concentrations. Meat Science, 83(1), 24–30. https://doi.org/10.1016/j.meatsci.2009.03.005 6 Z. González-Martínez et al. LWT 137 (2021) 110562 Nieto, G., Xiong, Y. L., Payne, F., & Castillo, M. (2014). Predicting frankfurters quality Rudibaugh, C. R. (1995). Meat emulsifying and processing system. In United States metrics using light backscatter. Journal of Food Engineering, 143, 132–138. https:// patent 08/205,588. doi.org/10.1016/j.jfoodeng.2014.06.027 Sebranek, J. G. (2003). MEAT | sausages and comminuted products. In Encyclopedia of Nieto, G., Xiong, Y. L., Payne, F., & Castillo, M. (2015). Light backscatter fiber optic Food sciences and nutrition (pp. 3785–3791). Elsevier. https://doi.org/10.1016/B0- sensor: A new tool for predicting the stability of pork emulsions containing 12-227055-X/00754-9. antioxidative potato protein hydrolysate. Meat Science, 100, 262–268. https://doi. Shan, B., Li, X., Pan, T., Zheng, L., Zhang, H., Guo, H., et al. (2014). Effect of shaddock org/10.1016/j.meatsci.2014.10.020 albedo addition on the properties of frankfurters. Journal of Food Science & Powers, R. G., Schack, W. R., & Anderson, J. E. (1987). Continuous vacuum grinding Technology, 52(7), 4572–4578. https://doi.org/10.1007/s13197-014-1467-7 method and apparatus. In United States patent 4,700,899. Tornberg, E. (2005). Effects of heat on meat proteins - implications on structure and Puolanne, E. (2010). Cooked sausages. In U. W. Oxford (Ed.), Handbook of meat quality of meat products. Meat Science, 70, 493–508. https://doi.org/10.1016/j. processing, products (pp. 313–325). https://doi.org/10.1002/9780813820897.ch17. meatsci.2004.11.021 Toldrá Fidel. Torres, V. (2016). Desarrollo de una tecnología óptica de control del grado de emulsificación en emulsiones cárnicas. Universitat Autónoma de Barcelona. 7