Introduction
In January 2024, the USDA Agricultural Marketing Service started a pilot program allowing official quality grading and certification using imaging technology for remote carcass grading (USDA, 2024). Thus, it is pivotal to understand current camera technologies and their ability to analyze meat quality attributes.
The most critical factors influencing beef consumers’ eating experience are tenderness, flavor, and juiciness. Essentially, if tenderness is at an adequate level, flavor becomes the major element that determines overall liking (Miller, 2020). Moreover, research has shown that intramuscular fat content can affect palatability and flavor (Smith et al., 1987). High intramuscular fat (marbling) has been associated with a positive impact on meat flavor (O’Quinn et al., 2012; Hunt et al., 2016). The average marbling score of carcasses from fed beef cattle has increased throughout years of selection (Gonzalez and Phelps, 2018). In 2022, the quality grade distribution of fed cattle harvested in the US was 8.7% USDA Prime, 70.4% USDA Choice, 17.0% USDA Select, and 4.84% other (USDA, 2023); thus, approximately 9% of US cattle could be considered highly marbled (slightly abundant or higher marbling scores). Accurate, consistent, and objective carcass assessment systems are essential for the meat industry to capture differences in marbling that affect eating quality and economic value.
As highly marbled carcasses become readily available, it is critical for processors to take advantage of technological modernizations to categorize and merchandise beef according to its intrinsic quality characteristics. Advances in machine learning through image analysis have resulted in the development of new objective carcass evaluation tools. Many of these cameras take an image of the longissimus thoracis muscle after the animal has been ribbed between the 12th and 13th ribs. These cameras objectively measure and predict characteristics such as marbling, fat content, ribeye area, and color from that image. Three existing and emerging cameras evaluated in this study included the Meat Image Japan (MIJ, Japan) camera, the MasterBeef (MB) (Australia) camera, and the VIAS VBG 2000 (E+V, Germany) camera. The E+V camera was developed in conjunction with the US Meat Animal Research Center, and it is currently used in the US, Canada, Australia, Mexico, and Russia (e+v Technology GmbH & Co. KG, 2021). The MIJ camera was developed in Japan by an R&D team led by Professor Keigo Kuchida and is used for grading carcasses in both Japan and Australia (Australian Wagyu Association, 2024). The MB camera was developed in Australia by Darren Hamblin and Peter Hobbs in collaboration with iScape, and now it is being used by 40 stakeholders, including South African Wagyu breeders (Condon, 2020). The MB system is a handheld smartphone with a high-resolution camera that works through a mobile application. Al/l cameras’ systems and software were developed to objectively assess beef carcass characteristics through image analysis of value-defining carcass attributes. The comparison of different camera technologies is of paramount importance due to varying levels of accuracy, camera and software adaptability, system scalability (line speed and volume), and cost. Therefore, supply chains should consider multiple elements when selecting camera technologies to objectively evaluate and grade beef carcasses.
Currently, the US beef industry relies heavily on USDA Quality Grades to predict the palatability of beef. There are no official references above Moderately Abundant marbling for highly marbled carcasses. In some cases, the intramuscular fat content of carcasses can exceed the existing marbling score scale (e.g., Extremely Abundant). Therefore, it is necessary to validate the relationship between new camera technologies and extracted percent intramuscular fat (%IMF) content in addition to USDA Marbling Scores. This study compares %IMF and marbling score predictability of the E+V, MIJ, and MB based on their unique fat measurements.
Materials and Methods
Camera data collection
A team of 3 USDA graders evaluated the left side of 176 F1 (50%) Wagyu carcasses after 24 h of postmortem chilling in a commercial beef processing facility. The same carcass sides were imaged with the E+V, MIJ, and MB cameras. Operators were trained according to manufacturer guidelines before imaging, and all measurements were taken by the same operator on the same day in a stationary rail. The attributes evaluated by each camera and an explanation of each variable are in Table 1.
Measurements and terminologies used in MB, MIJ, and E+V cameras
| Camera | Measurements | Definition |
|---|---|---|
| MasterBeef | Eye Muscle Area | Full eye muscle area measurement (cm2) |
| Fat Depth | Rib fat thickness (mm) | |
| Marble | Marbling score based on Beef Marbling Scoring | |
| Marbling Area | Area of fat taken from eye muscle area (cm2) | |
| Marbling Percent | Percent marbling area | |
| Marbling Total Perimeter | Total perimeter around meat sample (cm) | |
| Meat Area | Area of meat from eye muscle area (cm2) | |
| MIJ | Dma | Rib eye area (in2) |
| DMf | Fineness index | |
| DMp (%) | Marbling percent | |
| DMs | Marbling score based on Beef Marbling Scoring | |
| Taken at | Date and time | |
| E+V | Cam Marb Code | USDA marbling grade |
| Cam Marb Score | USDA numeric marbling score | |
| Cam Q Grade | USDA quality grade (Prime, Choice, Select) | |
| Cam REA | Ribeye area (in2) | |
| Cam Y Grade | Vision yield grade | |
| Grade Date | Date graded | |
| Grade Shrink | Difference between hot weight and grade weight (kg) | |
| Grade Weight | Weight when graded (kg) | |
| Hot Date | Date slaughtered | |
| Hot Weight | Weight when slaughtered (kg) | |
| Side ID | Carcass side |
Fat content
Following carcass grading and evaluation, a sliced sample of the left side longissimus thoracis approximately 5 mm thick was taken from each carcass and transported under refrigeration to the Center for Meat Safety and Quality at Colorado State University. Subcutaneous fat was removed, and all samples were then flash-frozen with liquid nitrogen and pulverized. Fat extraction was done as described by Swing et al. (2021). Briefly, one gram of the sample was weighed and placed in a 50-mL conical-bottom tube (VWR North American Cat. No. 8939-658; PA, USA), and 20 mL of 2:1 chloroform:methanol solution were added. Using a homogenizer (Scientific Pro homogenizer model 250; CT, USA), each sample was homogenized at the sixth speed for 30 s. Samples were then agitated on an orbital shaker (Burrel Corporation model 75; PA, USA) at room temperature for 20 min at 100 rpm. Homogenate was filtered using an ashless filter paper (Whatman #42 Cat. No. 1442-150; ME, UK) into prelabeled 20-mL glass vials. Four mL of 0.9% sodium chloride (NaCl) solution was added to each tube. Samples were covered with a plastic wrap (Parafilm M #HS234526C; WI, USA) and placed in a refrigerator at 4°C for approximately 16 h. The next day, the lower phase (with the lipid extract) was extracted using a glass pipette (Fisherbrand Cat. No. 13-678-20C; MA, USA) and placed in a preweighed labeled 16 × 50-mm scintillation vile (Fisherbrand 03-338B; MA, USA) that was placed into a 5.8-L glass desiccator (Pyrex model 3121-200; IL, USA) to dry for 2 h to evaporate residual chloroform and dried in a dry matter oven (GCA Precision model 26; IL, USA) at 100°C for approximately 16 h. Vials were allowed to cool at room temperature and weighed to calculate the %IMF of all carcass samples in triplicate.
Statistical analysis
Data were exported from all 3 camera systems into Microsoft Excel (Microsoft, Redmond, WA) and checked for missing values. Three data points (carcasses) were deleted because they did not have all 3 camera outputs, leading to a total number of 173 carcasses used for this experiment. Linear regression and descriptive statistics were performed using JMP (Statistical Discovery, NC, USA), analyzing the relationship between 2 variables using the fit Y by X function.
Results
This study aimed to describe the predictive potential of the E+V, MIJ, and MB beef grading cameras in estimating the %IMF in the longissimus thoracis muscle of F1 Wagyu cattle. Descriptive statistics are presented in Table 2. In this study, the E+V camera’s prediction of intramuscular fat is based on USDA Marbling Scores. The R2 for predictability of %IMF by the E+V camera was 0.6450, which was numerically higher than that of all other cameras’ and the USDA graders’ assessments. The MIJ camera assessed intramuscular fat based on marbling percent and marbling score. The estimates pertained to different intramuscular fat characteristics. Nevertheless, their predictability (R2) of %IMF did not differ (R2 = 0.5952). The MB estimated marbling (score), marbling area, and marbling percent. Although the measurements under consideration from the MB camera represented different aspects of intramuscular fat content within the longissimus muscle, the predictive capacity for %IMF of these measurements was numerically similar (Table 3) and lower than with the E+V system. Lastly, the USDA grader-assessed marbling scores resulted in an R2 = 0.6161 with extracted %IMF.
Descriptive statistics of percent intramuscular fat, E + V, MB, MIJ, and USDA grades.
| Statistic | %IMF* | E+V Score | MB | MIJ | USDA Grade | |||
|---|---|---|---|---|---|---|---|---|
| Area (cm2) | Marble | Percent | Percent | Score | ||||
| Mean | 14.37% | 828.78 | 57.31 | 7.73 | 22.79% | 13.48% | 3.97 | 795.08 |
| Standard Deviation | 4.40% | 139.05 | 16.69 | 2.06 | 6.65% | 6.61% | 1.94 | 160.60 |
| Min | 6.56% | 418.00 | 23.31 | 12.9 | 10.90% | 2.01% | 0.59 | 430.00 |
| Max | 30.64% | 1199.00 | 104.38 | 3.9 | 40.12% | 31.76% | 9.37 | 1099.00 |
%IMF - Percent intramuscular fat content.
Linear regression equations of the E+V, MB, MIJ, and USDA grader based on specific predictors
| Camera | Predictor | P Value | Root Mean Square Error | R2 | Equation |
|---|---|---|---|---|---|
| E+V | Marbling Score | <0.0001 | 0.0262 | 0.6450 | %IMF = −0.066833 + 0.0002541*Marbling Score |
| MasterBeef | Marble | <0.0001 | 0.0361 | 0.3269 | %IMF = 0.0494687 + 0.0121858*Marble |
| MasterBeef | Marbling Area | <0.0001 | 0.0360 | 0.3333 | %IMF = 0.0565526 + 0.0015214*Marbling Area |
| MasterBeef | Marbling Percent | <0.0001 | 0.0362 | 0.3249 | %IMF = 0.0578607 + 0.0037679*Marbling Percent |
| Meat Image Japan | Marbling Percent | <0.0001 | 0.0280 | 0.5952 | %IMF = 0.0744359 + 0.0051412*Marbling Percent |
| Meat Image Japan | Marbling Score | <0.0001 | 0.0280 | 0.5952 | %IMF = 0.0744359 + 0.017422*Marbling Score |
| USDA | Grade | <0.0001 | 0.0273 | 0.6161 | %IMF = −0.027206 + 0.000215*USDA Grade |
Discussion
Historically, the US has relied on highly skilled USDA graders to evaluate beef carcasses based on subjective assessment of carcass characteristics. New technologies involving electronic instruments, such as high-definition cameras and advanced image learning, can promote modern alternatives to beef carcass grading. These instruments have improved beef grading accuracy by minimizing subjectivity and reducing variability across USDA graders (Jang et al., 2017). The use of high-resolution cameras can facilitate the implementation of the USDA pilot for remote beef grading.
This study aimed to determine the accuracy of the E+V, MB, and MIJ cameras in predicting %IMF in the longissimus muscle of F1 Wagyu beef carcasses. The E+V camera demonstrated superior performance, yielding the highest R2 compared to the other 2 camera systems and the marbling scores assigned by USDA graders. Previous studies have confirmed the accuracy and precision of the E+V camera as a carcass grading tool (Allen, 2005; Dow et al., 2011). Schulz and Sundrum (2019) compared E+V measurements of %IMF at the 10th/11th rib with those obtained at the 12th/13th rib of the same carcass. Their findings were similar (R2 = 0.68) to the results found in this study (R2 = 0.6450) when predicting extractable %IMF using camera estimates of the marbling score (Schulz and Sundrum, 2019). The MIJ camera has previously been reported to accurately predict the quality of beef carcasses (Kuchida et al., 2000). However, recent studies suggest that the MIJ measurements can have variable precision when predicting %IMF (Stewart et al., 2021). Stewart et al. (2021) presented that the MIJ camera had an R2 = 0.5 to predict %IMF. In the current study, the MIJ had a higher coefficient of determination of R2 = 0.5952. Also in this study, the MB camera had the lowest R2 of all 3 cameras (Marbling area; R2 = 0.3333); therefore, further development and research are needed to improve the accuracy of the MB camera to establish the technology as a reliable grading tool for predicting %IMF. Additionally, with more data, researchers could better understand how the MB camera performs in different scenarios and under varying conditions.
The utilization of objective camera technologies provides an alternative for beef producers to assess carcass quality without human subjectivity. Additionally, camera technologies offer flexibility and efficiency because an individual who is not a USDA agent could be able to grade beef quality if trained properly. However, it is essential to consider that the USDA Quality Grades still hold significant value for the meat industry. In this study, subjective USDA grader assessments of marbling had the second highest R2 for %IMF compared to all the camera measurements (R2 = 0.6161). Undoubtedly, technology will continue to influence beef grading in the future. As technology continues to improve, its influence on beef quality grading will allow for revolutionizing improvements in efficiency and consistency.
Conclusion
This study compared the capability of E+V, MIJ, and MB to predict intramuscular fat in the longissimus thoracis. The E+V camera had the highest R2, followed by USDA graders, MIJ, and MB, respectively. New camera systems could serve as a tool for small and medium-sized meat processors for grading Wagyu-influenced beef carcasses. More profound knowledge of the ability of cameras to predict %IMF is needed to determine their accuracy. Additionally, the relationship between camera measurements and consumer acceptance should be further investigated.
Literature Cited
Allen, P. 2005. Evaluating video image analysis (VIA) systems for beef carcass classification [conference paper]. In: The science of beef quality. Eighth Annual Langford Food Industry Conference. pp. 9–11.
Dow, D. L., B. R. Wiegand, M. R. Ellersieck, and C. L. Lorenzen. 2011. Prediction of fat percentage within marbling score on beef longissimus muscle using 3 different fat determination methods. J. Anim. Sci. 89:1173–1179. doi: https://doi.org/10.2527/jas.2010-3382
e+v Technology GmbH & Co. KG. 2021. VGB 2000. https://www.eplusv.com/en/products/beef/vbg-2000/. (Accessed 15 October 2023).https://www.eplusv.com/en/products/beef/vbg-2000/
Gonzalez, J. M., and K. J. Phelps. 2018. United States beef quality as chronicled by the National Beef Quality Audits, Beef Consumer Satisfaction Projects, and National Beef Tenderness Surveys—A review. Asian Austral. J. Anim. 31:1036–1042. doi: https://doi.org/10.5713/ajas.18.0199
Hunt, M. R., J. F. Legako, T. T. N. Dinh, A. J. Garmyn, T. G. O’Quinn, C. H. Corbin, R. J. Rathmann, J. C. Brooks, and M. F. Miller. 2016. Assessment of volatile compounds, neutral and polar lipid fatty acids of 4 beef muscles from USDA Choice and Select graded carcasses and their relationships with consumer palatability scores and intramuscular fat content. Meat Sci. 116:91–101. doi: https://doi.org/10.1016/j.meatsci.2016.02.010
Jang, J. W., A. Ishdorj, D. P. Anderson, T. Purevjav, and G. Dahlke. 2017. Exploring the existence of grader bias in beef grading. Journal of Agricultural and Applied Economics 49:467–489. doi: https://doi.org/10.1017/aae.2017.9
MIJ Objective Carcase Measurement 2024. Australian Wagyu Association. https://www.wagyu.org.au/for-members/mij-objective-carcase-measurement. (Accessed 26 June 2024).https://www.wagyu.org.au/for-members/mij-objective-carcase-measurement
Kuchida, K., S. Kono, K. Konishi, L. Van Vleck, M. Suzuki, and S. Miyoshi. 2000. Prediction of crude fat content of longissimus muscle of beef using the ratio of fat area calculated from computer image analysis: Comparison of regression equations for prediction using different input devices at different stations. J. Anim. Sci. 78:799–803. doi: https://doi.org/10.2527/2000.784799x
Condon, J. 2020. Smart-phone based carcase grading camera shows big potential. Beef Central. https://www.beefcentral.com/news/smart-phone-based-carcase-grading-camera-shows-big-potential/. (Accessed 26 June 2024).https://www.beefcentral.com/news/smart-phone-based-carcase-grading-camera-shows-big-potential/
Miller, R. 2020. Drivers of consumer liking for beef, pork, and lamb: A review. Foods 9. Article 4. doi: https://doi.org/10.3390/foods9040428
O’Quinn, T. G., J. C., Brooks, R. J., Polkinghorne, A. J., Garmyn, B. J., Johnson, J. D., Starkey, R. J., Rathmann, and M. F. Miller. 2012. Consumer assessment of beef strip loin steaks of varying fat levels. J. Anim. Sci. 90:626–634. doi: https://doi.org/10.2527/jas.2011-4282
Schulz, L., and A. Sundrum. 2019. Assessing marbling scores of beef at the 10th rib vs. 12th rib of longissimus thoracis in the slaughter line using camera grading technology in Germany. Meat Sci. 152:116–120. doi: https://doi.org/10.1016/j.meatsci.2019.02.021
Smith, G. C., J. W. Savell, H. R. Cross, Z. l. Carpenter, C. E. Murphey, G. W. Davis, H. C. Abraham, F. C. Parrish, and B. W. Berry. 1987. Relationship of USDA quality grades to palatability of cooked beef. J. Food Quality 10:269–286. doi: https://doi.org/10.1111/j.1745-4557.1987.tb00819.x
Stewart, S. M., G. E. Gardner, A. Williams, D. W. Pethick, P. McGilchrist, and K. Kuchida. 2021. Association between visual marbling score and chemical intramuscular fat with camera marbling percentage in Australian beef carcasses. Meat Sci. 181:108369. doi: https://doi.org/10.1016/j.meatsci.2020.108369
Swing, C. J., T. W. Thompson, O. Guimaraes, I. Geornaras, T. E. Engle, K. E. Belk, M. N. Nair, L. Surati, and I. Utama. 2021. Nutritional composition of novel plant-based meat alternatives and traditional animal-based meats. Food Science & Nutrition 7:1–11.
USDA (US Department of Agriculture). 2023. USDA national steer and heifer estimated grading percent report. USDA market news. https://www.ams.usda.gov/mnreports/nw_ls196.txt. (Accessed 15 October 2023).https://www.ams.usda.gov/mnreports/nw_ls196.txt
USDA (US Department of Agriculture). 2024. USDA remote grading pilot for beef. USDA agricultural marketing service. https://www.ams.usda.gov/services/remote-beef-grading. (Accessed 22 February 2024).https://www.ams.usda.gov/services/remote-beef-grading