J. Anim Sci. 2008. 86:413-418. doi:10.2527/jas.2007-0095
© 2008 American Society of Animal Science
Using the near-infrared system to sort various beef middle and end muscle cuts into tenderness categories
D. M. Price,
G. G. Hilton,
D. L. VanOverbeke and
J. B. Morgan1
Oklahoma State University, Stillwater 74075
 |
Abstract
|
|---|
The objectives of this study were to determine the effectiveness of a visible-near-infrared (VIS-NIR) system to predict the ultimate tenderness rating of various beef muscles and conclude if a relationship exists between predicted LM shear force and tenderness of other subprimal cuts. Carcasses (n = 768) were scanned with the VIS-NIR system in 2 commercial beef-processing facilities. Carcasses were categorized based on their predicted 14-d LM slice shear force value. After carcass scanning, 100 carcasses were randomly selected based on their tenderness classification, and subprimals (ribeye rolls, clods, knuckles, top sirloins, inside rounds, and eye of rounds) were removed, vacuum-packaged, and transported to the Oklahoma State University Food and Agricultural Products Research Center, where 2.54-cm steaks (n = 6) were fabricated and stored in refrigerated conditions (1°C ± 1) and aged for 14 d. The center steak from right-side subprimals was designated for slice shear force (LM) or Warner-Bratzler shear force (all other subprimals) analysis. The remaining steaks were categorized based on predicted tenderness taken at 2 d postmortem with the VIS-NIR spectrophotometer and used in a consumer taste study. The test population of carcasses (n = 100) scanned in-plant predicted 27 carcasses as tender, 45 carcasses as intermediate, and 28 carcasses as tough. The VIS-NIR system correctly classified 26 of the 28 (92.9% accuracy) tough carcasses. Overall consumer satisfaction was greatest (P < 0.05) for steaks classified as tender and was intermediate compared with the steaks classified as tough. It was concluded that in-plant VIS-NIR scanning can properly identify and sort carcasses into tenderness groups, which may lead to the development of certified not-tough programs.
Key Words: beef • near-infrared • quality grade • tenderness
|
INTRODUCTION
|
|---|
Beef tenderness is a factor that must be incorporated in the quality grading process if true value-based marketing is to take place. In the US meat-marketing system, beef products leave the packing plant at about 3 d postmortem. It takes approximately 14 d for the beef products to reach the consumer. The beef industry needs an instrument that can scan fresh meat at 2 to 3 d postmortem and predict ultimate 14-d cooked-beef tenderness.
During the last decade, extensive research has been directed toward developing instruments for measuring beef palatability. An accurate, nondestructive method for online evaluation of tenderness continues to elude the beef industry. Research has shown that extracting wavelet textural features from ultrasonic elastogram images can predict Warner-Bratzler shear force (WBS) scores after aging for 2, 14, 28, and 42 d (R2 values ranged from 0.72 to 0.95; Huang et al., 1997
).
Light reflected in the visible region of the spectrum gives an objective measurement of the color of food objects, whereas light reflected in the visible-near-infrared (VIS-NIR) region contains information about food physical and chemical properties. In most studies, the VIS-NIR spectral scan and reference shear force tenderness values were acquired at the same point in time. Several studies have used scans acquired at early postmortem to predict tenderness after aging. The objectives of this study were to determine the effectiveness of the VIS-NIR system to predict the ultimate tenderness rating of various beef muscles from USDA Choice and Select carcasses and conclude if a relationship exists between predicted LM shear force value and muscle tenderness of other muscle cuts and to determine consumer perceptions of LM steaks, which were categorized using the VIS-NIR system as being tender, intermediate, or tough.
|
MATERIALS AND METHODS
|
|---|
Animal Care and Use Committee approval was not obtained for this study because the samples were obtained from federally inspected slaughter facilities.
Description of the Instrument
The VIS-NIR spectrometer used in this study (Field Spec Pro Jr., Analytical Spectral Devices Inc., Boulder, CO) was capable of collecting light in the visible and VIS-NIR regions (400 to 2,500 nm). A fiber-optic contact probe was used to transmit light reflected from the beef surface to 3 internal detectors. The detectors consisted of a silicon photodiode array, a thermoelectrically (TE) cooled indium gallium arsenide (In-GaAs), and a TE-cooled extended InGaAs to measure the 350 to 1,000-, 1,001 to 1,670-, and 1,671 to 2,500-nm wavelength domains, respectively. Light was supplied by a 20-W halogen light source and a diffuse reflection probe with 35° geometry with an effective measuring area of 1 mm2. The halogen lamp was powered by a feedback controller to stabilize illumination level.
Inside the instrument, a diffraction grating split the reflected light into narrow wavelength bands. A 512-channel silicon photodiode array was geometrically positioned to receive light within a narrow bandwidth (1.4 nm) in the region of 350 to 1,000 nm. The photodiodes converted the accumulated light to an electronic signal. The signal was digitized by and transferred to the computer. Spectral resolution in this region was 3 nm.
The 2 InGaAs detectors were the scanning type. They differed from the first sensor in that they measured wavelengths sequentially rather than simultaneously. Each sensor consists of a concave holographic grating and a single TE-cooled InGaAs detector. The gratings were mounted on a shaft that oscillates with a period of 200 ms (100 ms/scan). As the grating oscillates, the detector measures different wavelength bands. The resolution in these spectral regions was 30 nm. The spectrometer was carried in a backpack, with the laptop computer positioned ahead of the operator. The contact probe provided broadband light from an internal tungsten-halogen light source. The fiber-optic cable from the spectrometer terminated in a contact probe that projects broadband lighting and positions the cable to collect the light reflected from the beef surface.
Meat Samples
Beef carcasses (n = 768; 50% Low Choice and 50% Select) were randomly selected and scanned on d 2 postmortem with the VIS-NIR spectrometer in 1 of 2 regional packing plants (National Beef Inc., Liberal, KS and Sam Kane’s Inc., Corpus Christi, TX). Three spectra were collected at 3 locations near the lateral end of the ribeye in an effort to avoid connective tissue. The median of the 3 spectra was calculated and saved as the reflectance spectrum for that beef sample. Calculating the median avoided the effect of outliers related to a thick marbling spot or connective tissue. Optimization of integration time was accomplished by periodic placement of the contact probe on a white reference plate (Spectralon Diffuse Reflectance Targets, LabSphere Inc., North Sutton, NH). For a given scan, 10 spectra were collected consecutively and averaged to minimize electronic noise. Additionally, carcass grade data factors of fat thickness, ribeye area, lean maturity, skeletal maturity, marbling score, and quality grade, as evaluated and stamped by USDA grader, were collected. Hot carcass weight and carcass identification numbers were recorded from the plant tags.
Based on their 14-d predicted slice shear force (SSF) reading, as estimated by the VIS-NIR spectrometer (on d 2 postmortem), a LM tenderness grid was developed, and carcasses were categorized as being tender (<16 kg of SSF), intermediate (16 to 25 kg of SSF), or tough (>25 kg of SSF). Carcasses (n = 100) were randomly selected within tenderness category to be represented by approximately 25% tough and tender each, with the remaining selected carcasses (50%) being in the intermediate tenderness category.
After grade data collection, all carcasses were fabricated using standard techniques, and ribeye rolls [LM, Institutional Meat Purchase Specification (IMPS) #112A], clods (triceps brachii, IMPS #114), knuckles (rectus femoris, IMPS #167), top sirloins (gluteus medius, IMPS #184), inside rounds (semimembranosus, IMPS #168), and eyes of round (semitendinosus, IMPS #171C) were individually identified. After collection, all subprimals were vacuum-packaged, packed into boxes, and transported in refrigerated trucks from the respective packing plants to the Oklahoma State University (OSU) Food and Agricultural Products Research Center. At approximately 72 h postmortem, 2.54-cm steaks were progressively fabricated from the cranial end of each subprimal (Table 1), individually identified, and allowed to bloom for 30 min. The center steak (steak #3) from the right-side subprimal was designated for WBS or SSF (LM) analysis. All remaining steaks (n = 11) were aged for 14 d (1.0°C) and frozen (–20.0°C) until further analysis.
View this table:
[in this window]
[in a new window]
|
Table 1. Assignment of steaks from each subprimal (LM, triceps brachii, rectus femoris, gluteus medius, semimembranosus, semitendinosus)
|
|
Warner-Bratzler Shear Force Measurement
Center subprimal steaks were randomly distributed across each cooking date so that all quality grades and aging times were represented. One hundred steaks were allowed to temper daily at 4°C for 24 h before cooking. Steaks were broiled in an impingement oven (Lincoln Impinger, Model 1132-000-A, Lincoln Foodservice Products, Fort Wayne, IN) at 180°C to an internal temperature of 65°C; temperature was monitored using a Digi-Sense type T thermocouple (Model 91100-20, Cole-Palmer Instrument Company, Vernon Hills, IL). After cooking, steaks were allowed to cool to room temperature. A minimum of 6 cores (1.27-cm diam.) were removed parallel to muscle fiber orientation and sheared once, using a Warner-Bratzler head attached to an Instron Universal Testing Machine (Model 4502, Instron Corporation, Canton, MS). The Warner-Bratzler head moved at a crosshead speed of 200 mm/min. Peak load (kg) of each core was recorded by an IBM PS2 (Model 55 SX) using software provided by the Instron Corporation. Mean peak load (kg) was analyzed for each sample.
SSF Measurement
Ribeye steaks for SSF assessment were thawed and cooked in an identical manner to the other steaks in the investigation. Using the procedures outlined by Shackelford et al. (1999), a first cut was made 1 cm from the lateral end of the cooked steak. The second cut was made at 5 cm from the first cut. The slice shear sample was removed at an angle of 45° using a knife with 2 parallel blades separated by a 1-cm space. This procedure generated a cooked meat sample measuring 5 cm in length by 1 cm in thickness and 2.5 cm in width. This sample location was selected so that limited connective tissue would be located within the slice shear sample. Slice shear force was measured using a flat, blunt-end blade (slice shear attachment) attached to an Instron Universal Testing Machine (Instron Corporation). Greater SSF values indicated tougher beef.
Consumer Survey Testing
Ribeye rolls, IMPS #112A (National Association of Meat Purveyors, 1992), were cut into 2.54-cm-thick steaks. Steaks with an exposed LM of 
6 cm2 were eliminated from the investigation. The steaks were individually vacuum-packaged, aged for 14 d, frozen, and stored at –20.0°C. No attempt was made to categorize the steaks based on quality grade; however, it was not important how or why the meat was tough or tender but rather what the consumer response would be. After removing the center steaks from the right-side ribeye rolls, the remaining steaks were individually identified using color-coded adhesive tags, based on their respective predicted tenderness classification category: tender (red), intermediate (white), or tough (blue). With the exception of steak #3 from the right side, all remaining steaks (n = 11) were used for consumer evaluation. Thus, as many as 11 consumer steaks per ribeye roll could have been used.
Families (n = 117) were recruited from the Oklahoma City and Tulsa, Oklahoma, metropolitan areas to serve as consumers for the study. Specific demographics of the consumers were not obtained. The intent was not to project consumer targets but to obtain information so that an in-depth consumer study could be designed if the findings proved positive. Two steaks from each category were delivered to each household by OSU personnel. To account for sampling variation, the household was instructed as to the order in which the categories of steaks should be evaluated. There were 6 possible combinations in which the 3 color combinations could be ordered. These 6 possible combinations were randomly assigned among the 117 households. The steaks were identified to the consumers only by their color-coded labels. Therefore, consumers associated their likes and dislikes with a particular color-coding category.
Two participating adults from each family were given 3 wk to prepare and evaluate the steaks as they wished. An evaluation form for each steak was completed by each participant using a 23-point scale (National Live Stock and Meat Board, 1995). Evaluations were made concerning the satisfaction of the consumers with the product, as well as thawing methods, preparation method, and degree of doneness. At the time of delivery, OSU personnel presented the consumers with detailed instructions for completing the evaluation forms, both orally and in written form. The OSU personnel answered all questions and made certain that all consumers understood the requirements of their participation.
Model Development
Each VIS-NIR spectrum recorded had 2,150 data points or independent variables. Slice shear force was the dependent variable. A partial least squares (PLS) regression was employed to avoid overfitting. Partial least squares regression produces new features that are linear combinations of the original spectral data points, such that the new factors are not correlated and explain most of the variation in the dependent and independent variables (CAMO, 1998). Absorbance spectra in the region of 400 to 1,650 nm were used to predict SSF. All wavelengths from 400 to 1,650 were used in both PLS and principle component analysis. Any wavelengths that were considered outliers were eliminated from the study. Signal strength from reflections beyond 1,650 nm was determined to be below the threshold. The model was developed using Unscrambler software (CAMO, 1998). Cross-validation procedures (Esbenson, 2001) were employed to select the number of PLS factors to include in the models. A Savitsky-Golay cross-validation was used in the analysis of PLS factors. A total of 10 principle components were generated after completing a principle component analysis on the VIS-NIR spectra.
Our evaluation of system performance followed procedures described by Wheeler et al. (2002). They assessed performance of 3 instrumented tenderness prediction systems on the basis of progressive certification of steak sample tenderness in 10% certification increments. We classified any steaks with a 14-d SSF greater than 25 kg as tough and the rest as tender. In the description that follows, the observed values refer to the reference SSF values. Predicted values refer to the 14-d shear force predicted by the spectral reflectance system.
|
RESULTS AND DISCUSSION
|
|---|
As mentioned previously, 100% of the carcasses had small or slight marbling scores that would qualify them for USDA Choice and USDA Select quality grades, respectively (Table 2). The 2001 National Beef Quality Audit (McKenna et al., 2002) indicated that 83% of fed steer and heifer carcasses had slight or small marbling scores. Wulf and Page (2000) reported that in this narrow range of marbling scores, more precise methods for distinguishing palatable from unpalatable beef were needed.
Predicted 14-d ribeye SSF values were obtained at 2 d postmortem using the VIS-NIR system. The test population (n = 100) scanned in-plant using the VIS-NIR system predicted 27 carcasses as being tender, 45 carcasses as intermediate, and 28 carcasses as tough (Figure 1). In general, tough meat absorbed more light than tender meat. This result is in agreement with previous studies (Rødbotten et al., 2000; Leroy et al., 2003). Beyond 1,400 nm, the signal level is low due to water absorption. The reflectance (R) spectrum is converted to an absorbance spectrum by the log (1/R) transformation. This transformation is commonly employed to linearize the relationship between the concentration of an absorbing compound and the absorption spectrum (Hruschka, 2001).
View larger version (16K):
[in this window]
[in a new window]
|
Figure 1. Classification of LM into three 14-d tenderness classes based on the visible-near-infrared (VIS-NIR) tenderness classification system readings obtained in-plant on 2 d [tender = <16 kg of slice shear force (SSF); intermediate = 16 to 25 kg of SSF; tough = >25 kg of SSF].
|
|
The mean SSF values (14 d) were 15.4, 19.2, and 33.4 kg, respectively, for tender, intermediate, and tough. A low correlation coefficient between the observed and predicted SSF classification categories indicated that the VIS-NIR system did not predict specific tenderness values (i.e., tender and intermediate classifications) with high accuracy. However, the beef industry is more interested in identifying tough carcasses than predicting exact shear force values. A categorical model could be developed to classify carcasses into tender (SSF < 25 kg) or tough (SSF 
25 kg). By design, 28% of the carcasses tested in this study were tough. The VIS-NIR system correctly classified 26 of the 28 (92.9% accuracy) tough carcasses. Prior investigations involving the testing of various tenderness prediction systems automatically classified a majority (>93%) of carcasses as tender (Wheeler et al., 2002). This anomaly stems from the low number of tough carcasses in the sample populations.
When carcasses were segregated according to LM tenderness classification (tender = <16 kg of SSF; intermediate = 16 to 25 kg of SSF; or tough = >25 kg), differences in WBS of other muscles were not great. Longissimus muscle tenderness classification (tough vs. tender and intermediate) only correlated with WBS of triceps brachii (P < 0.001) and gluteus medius (P < 0.01). These findings were in agreement with Shackelford et al. (1997) in that they concluded that carcasses originating from Bos taurus and Bos indicus cattle generated tough triceps brachii and gluteus medius cuts when the same carcasses contained LM steaks displaying WBS values greater than 6.0 kg. These data suggest that genetic selection for improved LM tenderness may not have a large effect on tenderness of some muscles. Moreover, these data suggest that systems that accurately predict the tenderness of the LM of a carcass may not accurately predict the tenderness of other muscles.
Considering that it was our intention to include a tremendous amount of tenderness variation within all of the muscles evaluated (range in WBS exceeded 4 kg for all tested muscles; Table 3), variation in WBS of those muscles was not highly associated with variation in LM SSF classification using the VIS-NIR tenderness system. Thus, tenderness of each muscle will have to be predicted or methods to ensure the tenderness of each muscle will have to be used. Logically, the most efficient path to ensuring consistently tender beef would be to tenderize all cuts. Unfortunately, the beef industry has been reluctant to adopt more effective technologies (i.e., Ca-activated tenderization and enhancement) to improve tenderness. Instead, the meat industry has relied on segregation of carcasses and cuts into expected palatability outcome groups.
Similar to the tenderness classification of LM steaks, the VIS-NIR system was effective (P < 0.05) at placing less-desirable triceps brachii steaks into the tough classification category. It should be noted that the tender and intermediate tenderness classification groups were similar in their respective WBS means to the unsorted triceps brachii population. Percentages of individual tenderness classes accurately categorized as tender for gluteus medius classification were more similar to LM steaks than any of the other tested muscles (Table 4).
View this table:
[in this window]
[in a new window]
|
Table 4. Means and SEM of tenderness classification groups of rectus femoris, gluteus medius, semimembranosus, and semitendinosus muscles as predicted by the visible-near-infrared (VIS-NIR) system
|
|
Table 5 presents mean scores for consumer evaluations of beef LM steaks based on a 23-point scale. Overall satisfaction and tenderness ratings were greater (P < 0.05) for red and white LM steaks when compared with the steaks classified by the VIS-NIR system as being the toughest, the blue steaks. These data closely follow the SSF patterns of the paired LM steak counterparts in that the tough classified steaks were more easily found by the VIS-NIR system than differences between tender and intermediate steaks. No differences (P > 0.05) were observed between the different tenderness classifications for juiciness or flavor. The evaluation forms completed by each consumer provided information concerning thawing methods, preparation methods, and degree of doneness. However, no attempt was made to determine how these preparation methods may have influenced customer satisfaction.
View this table:
[in this window]
[in a new window]
|
Table 5. Mean scores for consumer palatability evaluations of beef LM steaks segmented according to the visible-near-infrared (VIS-NIR) system1
|
|
With the ability of consumers to discriminate among tenderness classification groups, it is possible that economic incentives may be used by beef processors, retailers, and food-service operators that can guarantee the elimination of tough beef from the supply. In an attempt to improve efficiency in processing facilities, carcasses that were predicted as tough could be accurately sorted onto specific rails in holding coolers, thus benefiting the fabrication process. In addition, identification of tough carcasses in-plant could benefit beef producers that produce cattle that generate tender carcasses, which could possibly lead to incentives for specific degrees of tenderness.
1 Corresponding author: bmorgan{at}okstate.edu
Received for publication February 13, 2007.
Accepted for publication October 18, 2007.
|
LITERATURE CITED
|
|---|
CAMO. 1998. The Unscrambler User Guide. Ver. 7.1 CAMO Inc., Corvallis, OR.
Esbenson, K. H. 2001. Multivariate data analysis—in practice. An introduction to multivariate data analysis and experimental design. 5th ed. CAMO Inc., Corvallis, OR.
Hruschka, W. R. 2001. Data Analysis: Wavelength selection methods. Pages 39–58 in Near-Infrared Technology in the Agricultural and Food Industries. P. Williams and K. Norris., ed. Am. Soc. Cereal Chem., St. Paul, MN.
Huang, Y., R. E. Lacey, L. L. Moore, R. K. Miller, A. D. Whittaker, and J. Ophir. 1997. Wavelet textural features from ultrasonic elastograms for meat quality prediction. Trans. ASAE 40:1741–1748.
Leroy, B., S. Lambotte, O. Dotreppe, H. Lecocq, L. Istasse, and A. Clinquart. 2003. Prediction of technological and organoleptic properties of beef Longissimus thoracis from near-infrared reflectance and transmission spectra. Meat Sci. 66:45–54.[CrossRef]
McKenna, D. R., D. L. Roebert, P. K. Bates, T. B. Schmidt, D. S. Hale, D. B. Griffin, J. W. Savell, J. C. Brooks, J. B. Morgan, T. H. Montgomery, K. E. Belk, and G. C. Smith. 2002. National Beef Quality Audit—2000: Survey of targeted cattle and carcass characteristics related to quality, quantity, and value of fed steers and heifers. J. Anim. Sci. 80:1212–1222.[Abstract/Free Full Text]
National Association of Meat Purveyors. 1992. The Meat Buyers Guide 3rd ed. Natl. Assoc. Meat Purveyors, Reston, VA.
National Live Stock and Meat Board. 1995. Beef Customer Satisfaction. A report to the industry. Natl. Live Stock Meat Board, Chicago, IL.
Rødbotten, R., B. N. Nilsen, and K. I. Hildrum. 2000. Prediction of beef quality attributes from early post mortem near infrared reflectance spectra. Food Chem. 69:427–436.[CrossRef]
Shackelford, S. D., T. L. Wheeler, and M. Koohmaraie. 1997. Tenderness classification of beef. I. Evaluation of beef longissimus shear force at 1 or 2 days postmortem as a predictor of aged beef tenderness. J. Anim. Sci. 75:2417–2422.[Abstract/Free Full Text]
Shackelford, S. D., T. L. Wheeler, and M. Koohmaraie. 1999. Tenderness classification of beef. II. Design and analysis of a system to measure beef longissimus shear force under commercial processing conditions. J. Anim. Sci. 77:1474–1481.[Abstract/Free Full Text]
Wheeler, T. L., D. Vote, J. M. Leheska, S. D. Shackelford, K. E. Belk, D. M. Wulf, B. L. Gwartney, and M. Koohmaraie. 2002. The efficacy of three objective systems for identifying beef cuts that can be guaranteed tender. J. Anim. Sci. 80:3315–3327.[Abstract/Free Full Text]
Wulf, D. M., and J. K. Page. 2000. Using measurements of muscle color, pH, and electrical impedance to augment the current USDA beef quality grading standards and improve the accuracy and precision of sorting carcasses into palatability groups. J. Anim. Sci. 78:2595–2607.[Abstract/Free Full Text]
Top
Abstract
INTRODUCTION
