TY - JOUR
AU - Smith, G. C.
AB - ABSTRACT This study was performed to validate previous equations and to develop and evaluate new regression equations for predicting lamb carcass fabrication yields using outputs from a lamb vision system-hot carcass component (LVS-HCC) and the lamb vision system-chilled carcass LM imaging component (LVS-CCC). Lamb carcasses (n = 149) were selected after slaughter, imaged hot using the LVS-HCC, and chilled for 24 to 48 h at −3 to 1°C. Chilled carcasses yield grades (YG) were assigned on-line by USDA graders and by expert USDA grading supervisors with unlimited time and access to the carcasses. Before fabrication, carcasses were ribbed between the 12th and 13th ribs and imaged using the LVS-CCC. Carcasses were fabricated into bone-in subprimal/primal cuts. Yields calculated included 1) saleable meat yield (SMY); 2) subprimal yield (SPY); and 3) fat yield (FY). On-line (whole-number) USDA YG accounted for 59, 58, and 64%; expert (whole-number) USDA YG explained 59, 59, and 65%; and expert (nearest-tenth) USDA YG accounted for 60, 60, and 67% of the observed variation in SMY, SPY, and FY, respectively. The best prediction equation developed in this trial using LVS-HCC output and hot carcass weight as independent variables explained 68, 62, and 74% of the variation in SMY, SPY, and FY, respectively. Addition of output from LVS-CCC improved predictive accuracy of the equations; the combined output equations explained 72 and 66% of the variability in SMY and SPY, respectively. Accuracy and repeatability of measurement of LM area made with the LVS-CCC also was assessed, and results suggested that use of LVS-CCC provided reasonably accurate (R2 = 0.59) and highly repeatable (repeatability = 0.98) measurements of LM area. Compared with USDA YG, use of the dual-component lamb vision system to predict cut yields of lamb carcasses improved accuracy and precision, suggesting that this system could have an application as an objective means for pricing carcasses in a value-based marketing system. Introduction The need in the lamb industry for an objective, accurate method of predicting red meat yield and monetary value of lamb carcasses has long been acknowledged. Brady et al. (2003) studied the ability of the lamb vision system (LVS; Research Management Systems U.S.A., Fort Collins, CO) to predict fabrication yields of lamb carcasses, and demonstrated that LVS predicted fabrication yield and monetary value of lamb carcasses more accurately than current pricing systems (based on live weight or hot carcass weight [HCW] and USDA yield grades). Purcell (1995) and Ward (1998) also reported that cutability differences are basically ignored on slaughter lambs pricing system. In order for the lamb industry to produce a leaner end product, there remains a need to develop an accurate method to predict saleable meat yields that will also serve as an objective method for assigning monetary value to lamb carcasses. Computer vision systems (CVS) are being used by some in the U.S. beef industry to estimate carcass yields. Steiner et al. (2000) employed both instrumentation and grader estimates in an on-line augmentation system to predict beef carcass yields. They reported that augmentation of the application of yield grades by USDA line graders with video image analysis (VIA) in real time significantly increased the accuracy with which subprimal yields were predicted (Steiner et al., 2000). In another study, Steiner et al. (2003) reported that VIA instrumentation could be used to assess LM area of beef carcasses with high levels of accuracy and repeatability. Therefore, the objectives of this study were to 1) validate the regression equations developed by Brady et al. (2003) to predict lamb carcass fabrication yields; 2) identify possible improvements to the accuracy and precision of those equations using both the LVS hot carcass component (LVS-HCC) and the chilled LM imaging system (LVS-CCC); and 3) assess the repeatability of LM area measurements using the LVS system. Experimental Procedures Lamb carcasses (n = 149) were selected by Colorado State University personnel at a commercial packing plant after slaughter but before entering the chilling coolers. During each of the five selection days of the study, carcasses were selected to fit a designed plan on the basis of gender class (ewe or wether), hot carcass weight (light = ≤29.48 kg; medium = 29.94 kg to 34.02 kg; or heavy = ≥34.47 kg), and USDA yield grade (1, 2, 3, 4, or 5; USDA, 1992). Within USDA yield grades, an approximately equal number of carcasses was selected to fill each of two muscling subclasses (light or heavy), which were included to ensure that cutability differences were not due to fatness alone. Lamb carcasses selected for inclusion in this study reflected the extreme range of variation in carcass traits experienced at the commercial facility each week (Table 1). Such variability ensured that the LVS (Research Management Systems U.S.A.) would be tested for accurate prediction of carcass cutability across the extreme range of differences in composition encountered in the present U.S. lamb population. Immediately after carcass selection (before chilling), carcass identification numbers and HCW were recorded, and images of each carcass were obtained using the LVS-HCC. Table 1. Number of carcasses included in the study stratified by gender class, hot carcass weight, and USDA yield grade (YG) category   Wether (n = 80)a    Ewe (n = 69)a      Expert (whole-number) USDA YG  Lightb  Mediumc  Heavyd  Lightb  Mediumc  Heavyd  Complete population  1  3  2  2  0  8  1  16  2  8  10  6  10  6  0  40  3  7  11  15  9  12  12  66  4  0  2  10  1  3  4  20  5  0  2  2  0  1  2  7  Total  18  27  35  20  30  19  149    Wether (n = 80)a    Ewe (n = 69)a      Expert (whole-number) USDA YG  Lightb  Mediumc  Heavyd  Lightb  Mediumc  Heavyd  Complete population  1  3  2  2  0  8  1  16  2  8  10  6  10  6  0  40  3  7  11  15  9  12  12  66  4  0  2  10  1  3  4  20  5  0  2  2  0  1  2  7  Total  18  27  35  20  30  19  149  a Lambs selected were approximately equal in number of light- vs. heavy-muscled lamb carcasses. b Light = ≤29.48 kg. c Medium = 29.94 to 34.02 kg. d Heavy = ≥34.47 kg. View Large Table 1. Number of carcasses included in the study stratified by gender class, hot carcass weight, and USDA yield grade (YG) category   Wether (n = 80)a    Ewe (n = 69)a      Expert (whole-number) USDA YG  Lightb  Mediumc  Heavyd  Lightb  Mediumc  Heavyd  Complete population  1  3  2  2  0  8  1  16  2  8  10  6  10  6  0  40  3  7  11  15  9  12  12  66  4  0  2  10  1  3  4  20  5  0  2  2  0  1  2  7  Total  18  27  35  20  30  19  149    Wether (n = 80)a    Ewe (n = 69)a      Expert (whole-number) USDA YG  Lightb  Mediumc  Heavyd  Lightb  Mediumc  Heavyd  Complete population  1  3  2  2  0  8  1  16  2  8  10  6  10  6  0  40  3  7  11  15  9  12  12  66  4  0  2  10  1  3  4  20  5  0  2  2  0  1  2  7  Total  18  27  35  20  30  19  149  a Lambs selected were approximately equal in number of light- vs. heavy-muscled lamb carcasses. b Light = ≤29.48 kg. c Medium = 29.94 to 34.02 kg. d Heavy = ≥34.47 kg. View Large The carcass assessment unit of the LVS consists of a stationary camera with built-in standardized lighting and a computer processor and monitor housed in a stainless steel cabinet. Following image acquisition, LVS software operates by 1) recording an image of a background, 2) recording an image of the carcass, and 3) subtracting the carcass image from the background image to provide a defined image of the carcass. In addition, LVS software recognizes all anatomical points that are needed to make a series of carcass measurements. Images were acquired on-line, with a speed of approximately 450 carcasses/h; carcasses were presented hanging on the rail (one trolley/carcass) and with the legs spread apart. Measurements made by the LVS software included, but was not limited to, carcass length, groin to right leg length, groin to left leg length, distance from groin to the end of the shank, red color score for shoulder, blue color score for shoulder, red color score for loin, blue color score for loin, distance between the two legs, groin area, carcass area measurements, total carcass width measurements, leg area measurements, leg width measurements, and groin angle measurements (Figure 1). These measurements were utilized as system output variables that describe the shape and size of the carcass, degree of muscularity, and relative proportions of fat and lean. Figure 1. View largeDownload slide Carcass outline with leg widths, groin angles, and maximum/minimum referenced widths. Figure 1. View largeDownload slide Carcass outline with leg widths, groin angles, and maximum/minimum referenced widths. On each selection day, carcasses were selected for scanning to ensure that an adequate number of carcasses would be available to fill the experimental design. Carcasses with slaughter defects that would preclude assignment of USDA quality and/or yield grades were not selected for the study. Test carcasses were then moved to the chilling cooler (−3 to 1°C) for 24 h of spray-chilling, except when it was not possible to perform fabrication of all selected carcasses the following day, which occurred twice during the study. On those occasions, carcasses were chilled for 48 h before fabrication. Following chilling, all carcasses were circulated past the grading stand, where, at normal chain speeds of approximately 480 carcasses/h, a USDA grader (employee of USDA-AMS) assessed and applied USDA quality and yield grades (USDA, 1992) to each carcass. After receiving on-line USDA grades, the selected carcasses were placed on static holding rails for further evaluation. An expert USDA grader (a field supervisor of USDA-AMS) then assigned and recorded USDA quality and yield grade factors (“Gold Standard” factors) and assigned a final expert USDA quality and yield grade to each carcass. During collection of expert grade data, the aid of a grading probe and unlimited time were allowed to maximize accuracy and precision of grade factor assignments. These “Gold Standard” factors and expert (whole number) USDA yield grades were used to determine which carcasses fit the project design. Each lamb carcass was then ribbed between the 12th and 13th ribs to expose the LM, and, following carcass grading but before fabrication, each carcass was scanned using the LVS-CCC in a manner such that three “triple-trigger” (camera head was positioned and three sequential images were obtained without moving or repositioning the camera head) and three “placement” (camera head was positioned and a single image was collected; the camera head was removed, repositioned, and a second image was collected, etc.) images were collected and recorded for each carcass as described by Steiner et al. (2003). The LVS-CCC consists of unit that records a digital image of the exposed surface of the interface of the 12th/13th ribs. The image was then processed to identify and measure LM area (LMA), to measure fat thickness opposite the LMA at three points, and to compute percentage of overall fat. Colorado State University personnel also measured LMA using a grid before fabrication (represented the “Gold Standard” LMA measurement). All fabrication steps were performed by experienced meat cutters (packing plant employees) supervised by Colorado State University personnel. The foresaddle and hindsaddle of each carcass were separated between the 12th and 13th ribs, weighed, and the sum of the two saddles was used to determine chilled carcass weight (CCW). The saddles were fabricated to generate primal cuts (shoulder, rack, loin, and leg). Each primal cut was weighed so that primal weights could be combined to allow carcass weight to be reconciled for each unit. Primal cut weights were reconciled by assuring that the combined weights of all bone-in cuts, lean trimmings, fat trimmings, and bone totaled at least 98% of initial primal weight. Subprimal/primal cuts generated from fabrication were closely trimmed of external fat (0.64 cm), and all subsequent parts were weighed and recorded. Bone-in subprimal cuts (these were not certified by USDA but closely approximated the appropriate Institutional Meat Purchasing Specifications [IMPS] description; USDA, 1996) included: square-cut shoulder (IMPS 207); neck; foreshank (IMPS 210); ribs, breast bones-off (IMPS 209A); split and chined rack (IMPS 204A); loin, short-cut, (IMPS 232A); leg (IMPS 233D); and hindshank (IMPS 233G). Data from eight carcasses were not included in the analysis, and reasons for exclusion from the analysis are shown in Table 2. Table 2. A summary of carcasses removed from the analyses Selection date  CSU ID No.  Reason  02-19-2003  160  Hot carcass weight < chilled carcass weight  02-19-2003  178  Hot carcass weight < chilled carcass weight  02-19-2003  195  Hot carcass weight < chilled carcass weight  03-03-2003  221  Not fabricated because of Yearling Mutton class  03-05-2003  236  <98% fabricated weight sum  03-05-2003  250  Not fabricated due to an injection-site lesion  03-05-2003  264  Missing fat trimmings from the leg  03-05-2003  285  Image and measurements not recorded properly  Selection date  CSU ID No.  Reason  02-19-2003  160  Hot carcass weight < chilled carcass weight  02-19-2003  178  Hot carcass weight < chilled carcass weight  02-19-2003  195  Hot carcass weight < chilled carcass weight  03-03-2003  221  Not fabricated because of Yearling Mutton class  03-05-2003  236  <98% fabricated weight sum  03-05-2003  250  Not fabricated due to an injection-site lesion  03-05-2003  264  Missing fat trimmings from the leg  03-05-2003  285  Image and measurements not recorded properly  View Large Table 2. A summary of carcasses removed from the analyses Selection date  CSU ID No.  Reason  02-19-2003  160  Hot carcass weight < chilled carcass weight  02-19-2003  178  Hot carcass weight < chilled carcass weight  02-19-2003  195  Hot carcass weight < chilled carcass weight  03-03-2003  221  Not fabricated because of Yearling Mutton class  03-05-2003  236  <98% fabricated weight sum  03-05-2003  250  Not fabricated due to an injection-site lesion  03-05-2003  264  Missing fat trimmings from the leg  03-05-2003  285  Image and measurements not recorded properly  Selection date  CSU ID No.  Reason  02-19-2003  160  Hot carcass weight < chilled carcass weight  02-19-2003  178  Hot carcass weight < chilled carcass weight  02-19-2003  195  Hot carcass weight < chilled carcass weight  03-03-2003  221  Not fabricated because of Yearling Mutton class  03-05-2003  236  <98% fabricated weight sum  03-05-2003  250  Not fabricated due to an injection-site lesion  03-05-2003  264  Missing fat trimmings from the leg  03-05-2003  285  Image and measurements not recorded properly  View Large Fabrication yields for bone-in cuts were calculated as a percentage of the CCW. The following yields were calculated for each carcass: “saleable meat yield” was the sum of weights of bone-in subprimal/primal cuts plus lean trimmings from the leg, loin, rack, shoulder, and thin cuts, expressed as a percentage of CCW; “subprimal yield” was the sum of weights of bone-in cuts from the leg, loin, rack, and shoulder expressed as a percentage of CCW; and “fat yield” was the sum of trimmable fat weights generated from production of subprimal/primal cuts and expressed as a percentage of CCW. Statistical Analyses All statistical analyses, including descriptive statistics, correlation, and regression analysis, were performed using PROC GLM and PROC REG of SAS (SAS Inst. Inc., Cary, NC). Multiple regression analysis was used to regress dependent carcass yield percentages on the independent variables of LVS-HCC output and HCW in an effort to develop an improved model and to validate the previously developed models (Brady et al., 2003) for the prediction of red meat yield. Regression equations for the prediction of weight of subprimal cuts from each primal were developed using linear methods that included LVS-HCC output variables and HCW. Forward and reverse stepwise regression methods were used to determine which independent variables were common and significant (α = 0.05) for each method of selection. Variables not selected by any of the three selection methods were excluded from the regression analysis, and the three selection methods were performed once more to build models for the various dependent carcass yield percentages. The root mean square error (RMSE) and predicted residual sum of squares (PRESS) statistics were computed to assess precision of red meat yield prediction models. Best-fit models were selected based on simplicity, R2, PRESS statistics, and RMSE values. Dependent carcass yield percents were also regressed on USDA on-line and expert yield grades, with yield grade serving as the sole independent variable in the model. The RMSE and PRESS statistics were calculated for each regression equation to determine the precision of the red meat yield model predictions. These regression equations were compared with the best-fit equations developed using LVS output and HCW as independent variables. Accuracy of LVS estimates of LMA for a given carcass were assessed by comparing each of the LVS estimates to “Gold Standard” LMA measurements collected by Colorado State University personnel using the regression procedure of SAS. Repeatability of LVS estimates of LMA were assessed first by determining the mean absolute differences between individual LVS-measured LMA and the average of those same measurements in each set of three images obtained from LVS. Secondly, repeatability for LVS-measured LMA was determined by evaluating variance components using the mixed-models procedure of SAS according to Montgomery (1997). Variance component estimation was conducted using the restricted maximum likelihood method. For each LMA, the proportion of total variance due to measurement replication was calculated as (σ2error)/(σ2carcass + σ2error). The model used to partition variance was as follows:  \[Y_{ij}\ =\ {\mu}\ +\ {\alpha}_{i}\ +\ {\varepsilon}_{ij}\] where Yij = ijth measured LMA, μ = the overall mean, αi = the random effect of the ith carcass, and εij = residual error. Results and Discussion Descriptive statistics for the 149 carcasses in the sample population are presented in Table 3. Inasmuch as the sample population was selected to represent an extreme range of variability in carcass traits present in a commercial facility, the large standard deviations for HCW, fat thickness, and USDA yield grade were due to intentional selection for carcass variation. Table 3. Descriptive statistics for carcass traits and USDA yield grades Item  n  Mean  SD  Minimum  Maximum  Hot carcass weight, kg  149  33.36  5.54  23.69  54.39  Chilled carcass weight, kg  149  33.17  5.50  23.60  54.03  Fat thickness, cm  149  0.70  0.23  0.13  1.37  Adjusted fat thickness, cm  149  0.72  0.23  0.13  1.42  Expert yield grades, nearest 10th  149  3.23  0.92  1.00  5.90  Expert yield grades, whole-number  149  2.73  0.98  1.00  5.00  On-line yield grades, whole-number  149  2.68  1.01  1.00  5.00  Item  n  Mean  SD  Minimum  Maximum  Hot carcass weight, kg  149  33.36  5.54  23.69  54.39  Chilled carcass weight, kg  149  33.17  5.50  23.60  54.03  Fat thickness, cm  149  0.70  0.23  0.13  1.37  Adjusted fat thickness, cm  149  0.72  0.23  0.13  1.42  Expert yield grades, nearest 10th  149  3.23  0.92  1.00  5.90  Expert yield grades, whole-number  149  2.73  0.98  1.00  5.00  On-line yield grades, whole-number  149  2.68  1.01  1.00  5.00  View Large Table 3. Descriptive statistics for carcass traits and USDA yield grades Item  n  Mean  SD  Minimum  Maximum  Hot carcass weight, kg  149  33.36  5.54  23.69  54.39  Chilled carcass weight, kg  149  33.17  5.50  23.60  54.03  Fat thickness, cm  149  0.70  0.23  0.13  1.37  Adjusted fat thickness, cm  149  0.72  0.23  0.13  1.42  Expert yield grades, nearest 10th  149  3.23  0.92  1.00  5.90  Expert yield grades, whole-number  149  2.73  0.98  1.00  5.00  On-line yield grades, whole-number  149  2.68  1.01  1.00  5.00  Item  n  Mean  SD  Minimum  Maximum  Hot carcass weight, kg  149  33.36  5.54  23.69  54.39  Chilled carcass weight, kg  149  33.17  5.50  23.60  54.03  Fat thickness, cm  149  0.70  0.23  0.13  1.37  Adjusted fat thickness, cm  149  0.72  0.23  0.13  1.42  Expert yield grades, nearest 10th  149  3.23  0.92  1.00  5.90  Expert yield grades, whole-number  149  2.73  0.98  1.00  5.00  On-line yield grades, whole-number  149  2.68  1.01  1.00  5.00  View Large Regression equations developed by Brady et al. (2003) were applied to carcasses in the current study data set, using the same independent variables but allowing coefficients to vary. Coefficients were allowed to differ from those of Brady et al. (2003) due to hardware and software adjustments incorporated into the LVS-HCC system during the time period between conducting the Brady et al. (2003) study and the current study; using these adjustments was necessary in order to improve and update the LVS-HCC. On-line (whole-number), expert (whole-number), and expert (nearest 10th) USDA yield grades were each regressed on carcass yields, and the values for R2, RMSE, and PRESS statistics for the regression equations are presented in Table 4. Expert (whole-number) USDA yield grades explained 59.1, 58.7, and 64.9% of the observed variability in saleable meat yields, subprimal yields, and fat yields, respectively. Table 4. Independent variables, R2, predicted residual sum of squares (PRESS), and root mean square error (RMSE) values for regression equations developed by Brady et al. (2003) to predict percent carcass yields using lamb vision system output plus hot carcass weight, and expert (whole-number and nearest 10th) and on-line (whole-number) USDA yield grades Dependent variable  R2  PRESS  RMSE  Variables in model (β-coefficient)a  Saleable meat yield, %b  0.5581  0.0946  0.0241  HCW (−0.00132)
 CxsLen (0.00043635)
 GrLftLeg (0.00071921)
 TaSum (−0.00000881)
 LwSum (0.00023675)
 GrAnSum (0.00013724)
 ShBi (0.78093)    0.5914  0.0777  0.0227  Expert (whole-number) USDA yield grade    0.6044  0.0754  0.0223  Expert (nearest 10th) USDA yield grade    0.5890  0.0781  0.0227  On-line (whole-number) USDA yield grade  Subprimal yield, %c  0.4588  0.0631  0.0199  HCW (−0.00137)
 CxsLen (0.00033006)
 LegGap (−0.00028752)
 ShBi (0.87778)
 TaSum (−0.00000386)
 LwSum (0.00007521)    0.5865  0.0443  0.0171  Expert (whole-number) USDA yield grade    0.5986  0.0430  0.0169  Expert (nearest 10th) USDA yield grade    0.5854  0.0444  0.0171  On-line (whole-number) USDA yield grade  Fat yield, %d  0.6155  0.0991  0.0249  HCW (0.00136)
 CxsLen (−0.00027979)
 ShRi (0.39424)
 TwSum (0.00005712)
 LwSum (−0.00001902)
 LaSum (−0.00009056)
 GrAnSum (−0.00020154)    0.6492  0.0820  0.0233  Expert (whole-number) USDA yield grade    0.6732  0.0767  0.0225  Expert (nearest 10th) USDA yield grade    0.6494  0.0820  0.0233  On-line (whole-number) USDA yield grade  Dependent variable  R2  PRESS  RMSE  Variables in model (β-coefficient)a  Saleable meat yield, %b  0.5581  0.0946  0.0241  HCW (−0.00132)
 CxsLen (0.00043635)
 GrLftLeg (0.00071921)
 TaSum (−0.00000881)
 LwSum (0.00023675)
 GrAnSum (0.00013724)
 ShBi (0.78093)    0.5914  0.0777  0.0227  Expert (whole-number) USDA yield grade    0.6044  0.0754  0.0223  Expert (nearest 10th) USDA yield grade    0.5890  0.0781  0.0227  On-line (whole-number) USDA yield grade  Subprimal yield, %c  0.4588  0.0631  0.0199  HCW (−0.00137)
 CxsLen (0.00033006)
 LegGap (−0.00028752)
 ShBi (0.87778)
 TaSum (−0.00000386)
 LwSum (0.00007521)    0.5865  0.0443  0.0171  Expert (whole-number) USDA yield grade    0.5986  0.0430  0.0169  Expert (nearest 10th) USDA yield grade    0.5854  0.0444  0.0171  On-line (whole-number) USDA yield grade  Fat yield, %d  0.6155  0.0991  0.0249  HCW (0.00136)
 CxsLen (−0.00027979)
 ShRi (0.39424)
 TwSum (0.00005712)
 LwSum (−0.00001902)
 LaSum (−0.00009056)
 GrAnSum (−0.00020154)    0.6492  0.0820  0.0233  Expert (whole-number) USDA yield grade    0.6732  0.0767  0.0225  Expert (nearest 10th) USDA yield grade    0.6494  0.0820  0.0233  On-line (whole-number) USDA yield grade  a HCW = hot carcass weight; CxsLen = carcass length; GrLftLeg = groin to left leg length; ShRi = red color score for shoulder (adjusted for intensity); ShBi = blue color score for shoulder (adjusted for intensity); LegGap = distance between the two legs; TaSum = sum of 20 total carcass area measurements; TwSum = sum of 20 total carcass width measurements; LaSum = sum of five leg area measurements; LwSum = sum of five leg width measurements; and GrAngSum = sum of five groin angle measurements. b Saleable meat yield = subprimal cuts and lean trim from the leg, loin, rack, shoulder, and thin cuts as a percentage of chilled side weight. c Subprimal yield = subprimal cuts from the leg, loin, rack, and shoulder as a percentage of chilled side weight. d Fat yield = percentage of chilled side weight of fat from the production of subprimal cuts. View Large Table 4. Independent variables, R2, predicted residual sum of squares (PRESS), and root mean square error (RMSE) values for regression equations developed by Brady et al. (2003) to predict percent carcass yields using lamb vision system output plus hot carcass weight, and expert (whole-number and nearest 10th) and on-line (whole-number) USDA yield grades Dependent variable  R2  PRESS  RMSE  Variables in model (β-coefficient)a  Saleable meat yield, %b  0.5581  0.0946  0.0241  HCW (−0.00132)
 CxsLen (0.00043635)
 GrLftLeg (0.00071921)
 TaSum (−0.00000881)
 LwSum (0.00023675)
 GrAnSum (0.00013724)
 ShBi (0.78093)    0.5914  0.0777  0.0227  Expert (whole-number) USDA yield grade    0.6044  0.0754  0.0223  Expert (nearest 10th) USDA yield grade    0.5890  0.0781  0.0227  On-line (whole-number) USDA yield grade  Subprimal yield, %c  0.4588  0.0631  0.0199  HCW (−0.00137)
 CxsLen (0.00033006)
 LegGap (−0.00028752)
 ShBi (0.87778)
 TaSum (−0.00000386)
 LwSum (0.00007521)    0.5865  0.0443  0.0171  Expert (whole-number) USDA yield grade    0.5986  0.0430  0.0169  Expert (nearest 10th) USDA yield grade    0.5854  0.0444  0.0171  On-line (whole-number) USDA yield grade  Fat yield, %d  0.6155  0.0991  0.0249  HCW (0.00136)
 CxsLen (−0.00027979)
 ShRi (0.39424)
 TwSum (0.00005712)
 LwSum (−0.00001902)
 LaSum (−0.00009056)
 GrAnSum (−0.00020154)    0.6492  0.0820  0.0233  Expert (whole-number) USDA yield grade    0.6732  0.0767  0.0225  Expert (nearest 10th) USDA yield grade    0.6494  0.0820  0.0233  On-line (whole-number) USDA yield grade  Dependent variable  R2  PRESS  RMSE  Variables in model (β-coefficient)a  Saleable meat yield, %b  0.5581  0.0946  0.0241  HCW (−0.00132)
 CxsLen (0.00043635)
 GrLftLeg (0.00071921)
 TaSum (−0.00000881)
 LwSum (0.00023675)
 GrAnSum (0.00013724)
 ShBi (0.78093)    0.5914  0.0777  0.0227  Expert (whole-number) USDA yield grade    0.6044  0.0754  0.0223  Expert (nearest 10th) USDA yield grade    0.5890  0.0781  0.0227  On-line (whole-number) USDA yield grade  Subprimal yield, %c  0.4588  0.0631  0.0199  HCW (−0.00137)
 CxsLen (0.00033006)
 LegGap (−0.00028752)
 ShBi (0.87778)
 TaSum (−0.00000386)
 LwSum (0.00007521)    0.5865  0.0443  0.0171  Expert (whole-number) USDA yield grade    0.5986  0.0430  0.0169  Expert (nearest 10th) USDA yield grade    0.5854  0.0444  0.0171  On-line (whole-number) USDA yield grade  Fat yield, %d  0.6155  0.0991  0.0249  HCW (0.00136)
 CxsLen (−0.00027979)
 ShRi (0.39424)
 TwSum (0.00005712)
 LwSum (−0.00001902)
 LaSum (−0.00009056)
 GrAnSum (−0.00020154)    0.6492  0.0820  0.0233  Expert (whole-number) USDA yield grade    0.6732  0.0767  0.0225  Expert (nearest 10th) USDA yield grade    0.6494  0.0820  0.0233  On-line (whole-number) USDA yield grade  a HCW = hot carcass weight; CxsLen = carcass length; GrLftLeg = groin to left leg length; ShRi = red color score for shoulder (adjusted for intensity); ShBi = blue color score for shoulder (adjusted for intensity); LegGap = distance between the two legs; TaSum = sum of 20 total carcass area measurements; TwSum = sum of 20 total carcass width measurements; LaSum = sum of five leg area measurements; LwSum = sum of five leg width measurements; and GrAngSum = sum of five groin angle measurements. b Saleable meat yield = subprimal cuts and lean trim from the leg, loin, rack, shoulder, and thin cuts as a percentage of chilled side weight. c Subprimal yield = subprimal cuts from the leg, loin, rack, and shoulder as a percentage of chilled side weight. d Fat yield = percentage of chilled side weight of fat from the production of subprimal cuts. View Large Equations developed by Brady et al. (2003) explained 55.8, 45.9 and 61.6% of the observed variability in saleable meat, subprimal, and fat yields, respectively (Table 4). Those values were lower than the values obtained by the USDA yield grades in the current study, but, when compared with the values from the previous study by Brady et al. (2003), the values obtained with the LVS-HCC were similar to those obtained by the USDA yield grades This indicated that LVS obtained similar results in both studies, but the USDA yield grades improved on explaining the amount of variability in lamb carcass cutability. As expected, the expert (nearest 10th) USDA yield grades explained slightly more of the observed variability in yields of saleable meat, subprimals, and fat than did on-line (whole-number) or expert (whole-number) USDA yield grades for the carcasses in this study (Table 4). Regression equations for expert (nearest 10th) USDA yield grades were accompanied by lower RMSE and PRESS values than those found for on-line or expert whole number USDA yield grades, suggesting greater precision in estimation of yields for saleable meat, subprimals, and fat. To further explore the use of the LVS-HCC output and to explain the variability in the shape of a lamb carcass as it changes due to fattening, new variables were developed using width measurements from LVS-HCC. These new variables included 1) a ratio between the elbow pocket width (minimum width measured in shoulder and rack) and the body wall width (maximum width measurement in the rack) and 2) a ratio between the loin width and shoulder width. These two variables, along with other LVS-HCC output variables, improved explanation of observed variation in percentage yields. The best-fit regression equation developed in the current study using the newly computed variables, along with HCW, explained 67.6, 61.9, and 73.8% of the observed variability in saleable meat, subprimal, and fat yields, respectively (Table 5), representing a significant improvement over use of USDA yield grades in the ability to predict percentage yields of lamb carcasses. Table 5. Independent variables, R2, root mean square error (RMSE), and predicted residual sum of squares (PRESS) statistics for best-fit regression equations developed to predict percent carcass yields using lamb vision system output plus hot carcass weight, developed in the present study and applied to the data from Brady et al. (2003)   Present study  Brady et al. (2003)  Dependent variable  R2  RMSE  PRESS  Variables in model (β-coefficient)a  R2  RMSE  PRESS  Variables in model (β-coefficient)a  Saleable meat yield, %b  0.676  0.0206  0.0686  HCW (−0.0021)
 CxsLen (0.000015)
 Sh:Ra ratio (0.26647)
 Tw ratio (−0.25329)
 GrRtLeg (0.000447)
 Lw1 (0.00064)
 ShBi (0.53452)  0.594  0.0216  0.1188  HCW (−0.0018)
 CxsLen (0.00030447)
 Sh:Ra ratio (0.21675)
 Tw ratio (−0.14012)
 GrRtLeg (0.0000093)
 Lw1 (0.00024)
 ShBi (1.50988)  Subprimal yield, %c  0.619  0.0168  0.0451  HCW (−0.00129)
 CxsLen (0.000087)
 Sh:Ra ratio (0.22107)
 Tw ratio (−0.21053)
 GrRtLeg (0.000273)
 Lw1 (0.000311)
 ShBi (0.6423)  0.548  0.021  0.112  HCW (−0.00166)
 CxsLen (0.0003042)
 Sh:Ra ratio (0.21016)
 Tw ratio (−0.14084)
 GrRtLeg (−0.000008)
 Lw1 (−0.000102244)
 ShBi (1.11035)  Fat yield %d  0.738  0.0205  0.0680  HCW (0.00267)
 CxsLen (−0.00016)
 Sh:Ra ratio (−0.27358)
 Tw ratio (0.2555)
 GrnAng1 (0.00226)
 Lw1 (−0.00076288)
 ShBi (−0.46843)  0.7542  0.0237  0.142  HCW (0.00287)
 CxsLen (−0.0003202)
 Sh:Ra ratio (−0.29372)
 Tw ratio (0.19025)
 GrnAng1 (0.00227)
 Lw1 (−0.00085889)
 ShBi (−1.56582)    Present study  Brady et al. (2003)  Dependent variable  R2  RMSE  PRESS  Variables in model (β-coefficient)a  R2  RMSE  PRESS  Variables in model (β-coefficient)a  Saleable meat yield, %b  0.676  0.0206  0.0686  HCW (−0.0021)
 CxsLen (0.000015)
 Sh:Ra ratio (0.26647)
 Tw ratio (−0.25329)
 GrRtLeg (0.000447)
 Lw1 (0.00064)
 ShBi (0.53452)  0.594  0.0216  0.1188  HCW (−0.0018)
 CxsLen (0.00030447)
 Sh:Ra ratio (0.21675)
 Tw ratio (−0.14012)
 GrRtLeg (0.0000093)
 Lw1 (0.00024)
 ShBi (1.50988)  Subprimal yield, %c  0.619  0.0168  0.0451  HCW (−0.00129)
 CxsLen (0.000087)
 Sh:Ra ratio (0.22107)
 Tw ratio (−0.21053)
 GrRtLeg (0.000273)
 Lw1 (0.000311)
 ShBi (0.6423)  0.548  0.021  0.112  HCW (−0.00166)
 CxsLen (0.0003042)
 Sh:Ra ratio (0.21016)
 Tw ratio (−0.14084)
 GrRtLeg (−0.000008)
 Lw1 (−0.000102244)
 ShBi (1.11035)  Fat yield %d  0.738  0.0205  0.0680  HCW (0.00267)
 CxsLen (−0.00016)
 Sh:Ra ratio (−0.27358)
 Tw ratio (0.2555)
 GrnAng1 (0.00226)
 Lw1 (−0.00076288)
 ShBi (−0.46843)  0.7542  0.0237  0.142  HCW (0.00287)
 CxsLen (−0.0003202)
 Sh:Ra ratio (−0.29372)
 Tw ratio (0.19025)
 GrnAng1 (0.00227)
 Lw1 (−0.00085889)
 ShBi (−1.56582)  a HCW = hot carcass weight; CxsLen = carcass length; GrRtLeg = groin to right leg length; ShBi = blue color score for shoulder (adjusted for intensity); Sh:Ra ratio = ratio of the maximum rack width and maximum shoulder width; Tw ratio = ratio of the minimum body (shoulder, rack, loin) and the maximum body width; Lw1 = leg width measurement closest to the groin; and GrnAng1 = first groin angle measurement. b Saleable meat yield = subprimal cuts and lean trim from the leg, loin, rack, shoulder, and thin cuts as a percentage of chilled weight. c Subprimal yield = subprimal cuts from the leg, loin, rack, and shoulder as a percentage of chilled side weight. d Fat yield = percentage of chilled side weight of fat from the production of subprimal cuts. View Large Table 5. Independent variables, R2, root mean square error (RMSE), and predicted residual sum of squares (PRESS) statistics for best-fit regression equations developed to predict percent carcass yields using lamb vision system output plus hot carcass weight, developed in the present study and applied to the data from Brady et al. (2003)   Present study  Brady et al. (2003)  Dependent variable  R2  RMSE  PRESS  Variables in model (β-coefficient)a  R2  RMSE  PRESS  Variables in model (β-coefficient)a  Saleable meat yield, %b  0.676  0.0206  0.0686  HCW (−0.0021)
 CxsLen (0.000015)
 Sh:Ra ratio (0.26647)
 Tw ratio (−0.25329)
 GrRtLeg (0.000447)
 Lw1 (0.00064)
 ShBi (0.53452)  0.594  0.0216  0.1188  HCW (−0.0018)
 CxsLen (0.00030447)
 Sh:Ra ratio (0.21675)
 Tw ratio (−0.14012)
 GrRtLeg (0.0000093)
 Lw1 (0.00024)
 ShBi (1.50988)  Subprimal yield, %c  0.619  0.0168  0.0451  HCW (−0.00129)
 CxsLen (0.000087)
 Sh:Ra ratio (0.22107)
 Tw ratio (−0.21053)
 GrRtLeg (0.000273)
 Lw1 (0.000311)
 ShBi (0.6423)  0.548  0.021  0.112  HCW (−0.00166)
 CxsLen (0.0003042)
 Sh:Ra ratio (0.21016)
 Tw ratio (−0.14084)
 GrRtLeg (−0.000008)
 Lw1 (−0.000102244)
 ShBi (1.11035)  Fat yield %d  0.738  0.0205  0.0680  HCW (0.00267)
 CxsLen (−0.00016)
 Sh:Ra ratio (−0.27358)
 Tw ratio (0.2555)
 GrnAng1 (0.00226)
 Lw1 (−0.00076288)
 ShBi (−0.46843)  0.7542  0.0237  0.142  HCW (0.00287)
 CxsLen (−0.0003202)
 Sh:Ra ratio (−0.29372)
 Tw ratio (0.19025)
 GrnAng1 (0.00227)
 Lw1 (−0.00085889)
 ShBi (−1.56582)    Present study  Brady et al. (2003)  Dependent variable  R2  RMSE  PRESS  Variables in model (β-coefficient)a  R2  RMSE  PRESS  Variables in model (β-coefficient)a  Saleable meat yield, %b  0.676  0.0206  0.0686  HCW (−0.0021)
 CxsLen (0.000015)
 Sh:Ra ratio (0.26647)
 Tw ratio (−0.25329)
 GrRtLeg (0.000447)
 Lw1 (0.00064)
 ShBi (0.53452)  0.594  0.0216  0.1188  HCW (−0.0018)
 CxsLen (0.00030447)
 Sh:Ra ratio (0.21675)
 Tw ratio (−0.14012)
 GrRtLeg (0.0000093)
 Lw1 (0.00024)
 ShBi (1.50988)  Subprimal yield, %c  0.619  0.0168  0.0451  HCW (−0.00129)
 CxsLen (0.000087)
 Sh:Ra ratio (0.22107)
 Tw ratio (−0.21053)
 GrRtLeg (0.000273)
 Lw1 (0.000311)
 ShBi (0.6423)  0.548  0.021  0.112  HCW (−0.00166)
 CxsLen (0.0003042)
 Sh:Ra ratio (0.21016)
 Tw ratio (−0.14084)
 GrRtLeg (−0.000008)
 Lw1 (−0.000102244)
 ShBi (1.11035)  Fat yield %d  0.738  0.0205  0.0680  HCW (0.00267)
 CxsLen (−0.00016)
 Sh:Ra ratio (−0.27358)
 Tw ratio (0.2555)
 GrnAng1 (0.00226)
 Lw1 (−0.00076288)
 ShBi (−0.46843)  0.7542  0.0237  0.142  HCW (0.00287)
 CxsLen (−0.0003202)
 Sh:Ra ratio (−0.29372)
 Tw ratio (0.19025)
 GrnAng1 (0.00227)
 Lw1 (−0.00085889)
 ShBi (−1.56582)  a HCW = hot carcass weight; CxsLen = carcass length; GrRtLeg = groin to right leg length; ShBi = blue color score for shoulder (adjusted for intensity); Sh:Ra ratio = ratio of the maximum rack width and maximum shoulder width; Tw ratio = ratio of the minimum body (shoulder, rack, loin) and the maximum body width; Lw1 = leg width measurement closest to the groin; and GrnAng1 = first groin angle measurement. b Saleable meat yield = subprimal cuts and lean trim from the leg, loin, rack, shoulder, and thin cuts as a percentage of chilled weight. c Subprimal yield = subprimal cuts from the leg, loin, rack, and shoulder as a percentage of chilled side weight. d Fat yield = percentage of chilled side weight of fat from the production of subprimal cuts. View Large To verify the validity of this newly developed equation, it was applied to data collected by Brady et al. (2003), and comparable results were achieved (Table 5). Differences between coefficients for the equations developed in the two studies were, again, explained by adjustments made to the hardware and software of the LVS-HCC between the times of the two trials. The slight difference in results from the Brady et al. (2003) study and the current study for R2, RMSE, and PRESS statistics is most likely explained by differences among lamb carcasses in the two populations. The Brady et al. (2003) population was more evenly distributed across all five USDA yield grades, whereas the population of the current study contained disproportionately more carcasses of yield grades 2 and 3, which is more representative of U.S. lamb carcasses. Using a combination of outputs from LVS-HCC and LVS-CCC improved predictive accuracy of yield estimation. Regression equations developed using a combination of outputs plus HCW and percent fat (PF), or alternatively, LMA plus PF explained 71.9% of the variation in saleable meat yield (Table 6). Descriptive statistics for those measurements are presented in Table 7. Table 6. Independent variables, R2, root mean square error (RMSE), predicted residual sum of squares (PRESS), and partial R2 for regression equations using lamb vision system-hot carcass component (LVS-HCC) factors, hot carcass weight (HCW), and LM area (LMA) and/or percent fat (PF) measurements obtained with lamb vision system-chilled carcass component (LVS-CCC), to predict saleable meat and subprimal yields from lamb carcasses Item  R2  RMSE  PRESS  Variables in model (partial R2)a  LVS-HCC, HCW, and LMA      Saleable meat yield, %b  0.6778  0.0205  0.0675  HCW (0.171), CxsLen (0.287),
 Sh:Ra ratio (0.064), Tw ratio (0.072),
 GrRtLeg (0.047), Lw1 (0.026),
 LMA (0.011)      Subprimal yield, %c  0.6411  0.0163  0.0425  HCW (0.126), Sh:Ra ratio (0.176),
 Tw ratio (0.192), GrAng1 (0.092),
 LMA (0.036), ShBi (0.009),
 Lw1 (0.008)      Fat yield, %d  0.7413  0.0241  0.0592  HCW (0.203), CxsLen (0.329),
 Sh/Ra ratio (0.054), Tw ratio (0.031),
 GrRtLeg (0.042), Lw1 (0.006),
 LMA (0.022)  LVS-HCC, HCW, and PF      Saleable meat yield, %b  0.7189  0.0192  0.0587  PF (0.526), HCW (0.065),
 CxsLen (0.068), Sh:Ra ratio (0.020),
 Tw ratio (0.010), GrRtLeg (0.025),
 Lw1 (0.006)      Subprimal yield, %c  0.6562  0.0159  0.0401  PF (0.468), HCW (0.063),
 Sh:Ra ratio (0.43), Tw ratio (0.020),
 GrAng1 (0.048), ShBi (0.015)      Fat yield, %d  0.7952  0.0182  0.0527  PF (0.573), HCW (0.081),
 CxsLen (0.084), Sh:Ra ratio (0.014),
 Tw ratio (0.007), GrRtLeg (0.025),
 Lw1 (0.007), ShBi (0.005)  LVS-HCC, HCW, LMA, and PF      Saleable meat yield, %b  0.7189  0.0192  0.0587  PF (0.526), HCW (0.065),
 CxsLen (0.068), Sh:Ra ratio (0.020),
 Tw ratio (0.010), GrRtLeg (0.025),
 Lw1 (0.006)      Subprimal yield, %c  0.6562  0.0159  0.0401  PF (0.468), HCW (0.063),
 Sh:Ra ratio (0.043), Tw ratio (0.020),
 GrAng1 (0.048), ShBi (0.014)      Fat yield, %d  0.7952  0.0182  0.0527  PF (0.573), HCW (0.081),
 CxsLen (0.084), Sh:Ra ratio (0.014),
 Tw ratio (0.007), GrRtLeg (0.025),
 Lw1 (0.007), ShBi (0.005)  Item  R2  RMSE  PRESS  Variables in model (partial R2)a  LVS-HCC, HCW, and LMA      Saleable meat yield, %b  0.6778  0.0205  0.0675  HCW (0.171), CxsLen (0.287),
 Sh:Ra ratio (0.064), Tw ratio (0.072),
 GrRtLeg (0.047), Lw1 (0.026),
 LMA (0.011)      Subprimal yield, %c  0.6411  0.0163  0.0425  HCW (0.126), Sh:Ra ratio (0.176),
 Tw ratio (0.192), GrAng1 (0.092),
 LMA (0.036), ShBi (0.009),
 Lw1 (0.008)      Fat yield, %d  0.7413  0.0241  0.0592  HCW (0.203), CxsLen (0.329),
 Sh/Ra ratio (0.054), Tw ratio (0.031),
 GrRtLeg (0.042), Lw1 (0.006),
 LMA (0.022)  LVS-HCC, HCW, and PF      Saleable meat yield, %b  0.7189  0.0192  0.0587  PF (0.526), HCW (0.065),
 CxsLen (0.068), Sh:Ra ratio (0.020),
 Tw ratio (0.010), GrRtLeg (0.025),
 Lw1 (0.006)      Subprimal yield, %c  0.6562  0.0159  0.0401  PF (0.468), HCW (0.063),
 Sh:Ra ratio (0.43), Tw ratio (0.020),
 GrAng1 (0.048), ShBi (0.015)      Fat yield, %d  0.7952  0.0182  0.0527  PF (0.573), HCW (0.081),
 CxsLen (0.084), Sh:Ra ratio (0.014),
 Tw ratio (0.007), GrRtLeg (0.025),
 Lw1 (0.007), ShBi (0.005)  LVS-HCC, HCW, LMA, and PF      Saleable meat yield, %b  0.7189  0.0192  0.0587  PF (0.526), HCW (0.065),
 CxsLen (0.068), Sh:Ra ratio (0.020),
 Tw ratio (0.010), GrRtLeg (0.025),
 Lw1 (0.006)      Subprimal yield, %c  0.6562  0.0159  0.0401  PF (0.468), HCW (0.063),
 Sh:Ra ratio (0.043), Tw ratio (0.020),
 GrAng1 (0.048), ShBi (0.014)      Fat yield, %d  0.7952  0.0182  0.0527  PF (0.573), HCW (0.081),
 CxsLen (0.084), Sh:Ra ratio (0.014),
 Tw ratio (0.007), GrRtLeg (0.025),
 Lw1 (0.007), ShBi (0.005)  a HCW = hot carcass weight; CxsLen = carcass length; GrRtLeg = groin to left leg length; ShRi = red color score for shoulder (adjusted for intensity); ShBi = blue color score for shoulder (adjusted for intensity); LMA = longissimus muscle area obtained from LVS-CCC; PF = overall percent fat in the 12th-/13th-rib interface obtained from VIA output; Sh:Ra ratio = ratio of the maximum rack width and maximum shoulder width; Tw ratio = ratio of the minimum body width (shoulder, rack, loin) and the maximum body width; Lw1 = leg width measurement closest to the groin; and GrAng1 = first groin angle measurement. b Saleable meat yield = subprimal cuts and lean trim from the leg, loin, rack, shoulder, and thin cuts as a percentage of chilled carcass weight. c Subprimal yield = subprimal cuts from the leg, loin, rack, and shoulder as a percentage of carcass side weight. d Fat yield = percentage of chilled side weight of fat from the production of subprimal cuts. View Large Table 6. Independent variables, R2, root mean square error (RMSE), predicted residual sum of squares (PRESS), and partial R2 for regression equations using lamb vision system-hot carcass component (LVS-HCC) factors, hot carcass weight (HCW), and LM area (LMA) and/or percent fat (PF) measurements obtained with lamb vision system-chilled carcass component (LVS-CCC), to predict saleable meat and subprimal yields from lamb carcasses Item  R2  RMSE  PRESS  Variables in model (partial R2)a  LVS-HCC, HCW, and LMA      Saleable meat yield, %b  0.6778  0.0205  0.0675  HCW (0.171), CxsLen (0.287),
 Sh:Ra ratio (0.064), Tw ratio (0.072),
 GrRtLeg (0.047), Lw1 (0.026),
 LMA (0.011)      Subprimal yield, %c  0.6411  0.0163  0.0425  HCW (0.126), Sh:Ra ratio (0.176),
 Tw ratio (0.192), GrAng1 (0.092),
 LMA (0.036), ShBi (0.009),
 Lw1 (0.008)      Fat yield, %d  0.7413  0.0241  0.0592  HCW (0.203), CxsLen (0.329),
 Sh/Ra ratio (0.054), Tw ratio (0.031),
 GrRtLeg (0.042), Lw1 (0.006),
 LMA (0.022)  LVS-HCC, HCW, and PF      Saleable meat yield, %b  0.7189  0.0192  0.0587  PF (0.526), HCW (0.065),
 CxsLen (0.068), Sh:Ra ratio (0.020),
 Tw ratio (0.010), GrRtLeg (0.025),
 Lw1 (0.006)      Subprimal yield, %c  0.6562  0.0159  0.0401  PF (0.468), HCW (0.063),
 Sh:Ra ratio (0.43), Tw ratio (0.020),
 GrAng1 (0.048), ShBi (0.015)      Fat yield, %d  0.7952  0.0182  0.0527  PF (0.573), HCW (0.081),
 CxsLen (0.084), Sh:Ra ratio (0.014),
 Tw ratio (0.007), GrRtLeg (0.025),
 Lw1 (0.007), ShBi (0.005)  LVS-HCC, HCW, LMA, and PF      Saleable meat yield, %b  0.7189  0.0192  0.0587  PF (0.526), HCW (0.065),
 CxsLen (0.068), Sh:Ra ratio (0.020),
 Tw ratio (0.010), GrRtLeg (0.025),
 Lw1 (0.006)      Subprimal yield, %c  0.6562  0.0159  0.0401  PF (0.468), HCW (0.063),
 Sh:Ra ratio (0.043), Tw ratio (0.020),
 GrAng1 (0.048), ShBi (0.014)      Fat yield, %d  0.7952  0.0182  0.0527  PF (0.573), HCW (0.081),
 CxsLen (0.084), Sh:Ra ratio (0.014),
 Tw ratio (0.007), GrRtLeg (0.025),
 Lw1 (0.007), ShBi (0.005)  Item  R2  RMSE  PRESS  Variables in model (partial R2)a  LVS-HCC, HCW, and LMA      Saleable meat yield, %b  0.6778  0.0205  0.0675  HCW (0.171), CxsLen (0.287),
 Sh:Ra ratio (0.064), Tw ratio (0.072),
 GrRtLeg (0.047), Lw1 (0.026),
 LMA (0.011)      Subprimal yield, %c  0.6411  0.0163  0.0425  HCW (0.126), Sh:Ra ratio (0.176),
 Tw ratio (0.192), GrAng1 (0.092),
 LMA (0.036), ShBi (0.009),
 Lw1 (0.008)      Fat yield, %d  0.7413  0.0241  0.0592  HCW (0.203), CxsLen (0.329),
 Sh/Ra ratio (0.054), Tw ratio (0.031),
 GrRtLeg (0.042), Lw1 (0.006),
 LMA (0.022)  LVS-HCC, HCW, and PF      Saleable meat yield, %b  0.7189  0.0192  0.0587  PF (0.526), HCW (0.065),
 CxsLen (0.068), Sh:Ra ratio (0.020),
 Tw ratio (0.010), GrRtLeg (0.025),
 Lw1 (0.006)      Subprimal yield, %c  0.6562  0.0159  0.0401  PF (0.468), HCW (0.063),
 Sh:Ra ratio (0.43), Tw ratio (0.020),
 GrAng1 (0.048), ShBi (0.015)      Fat yield, %d  0.7952  0.0182  0.0527  PF (0.573), HCW (0.081),
 CxsLen (0.084), Sh:Ra ratio (0.014),
 Tw ratio (0.007), GrRtLeg (0.025),
 Lw1 (0.007), ShBi (0.005)  LVS-HCC, HCW, LMA, and PF      Saleable meat yield, %b  0.7189  0.0192  0.0587  PF (0.526), HCW (0.065),
 CxsLen (0.068), Sh:Ra ratio (0.020),
 Tw ratio (0.010), GrRtLeg (0.025),
 Lw1 (0.006)      Subprimal yield, %c  0.6562  0.0159  0.0401  PF (0.468), HCW (0.063),
 Sh:Ra ratio (0.043), Tw ratio (0.020),
 GrAng1 (0.048), ShBi (0.014)      Fat yield, %d  0.7952  0.0182  0.0527  PF (0.573), HCW (0.081),
 CxsLen (0.084), Sh:Ra ratio (0.014),
 Tw ratio (0.007), GrRtLeg (0.025),
 Lw1 (0.007), ShBi (0.005)  a HCW = hot carcass weight; CxsLen = carcass length; GrRtLeg = groin to left leg length; ShRi = red color score for shoulder (adjusted for intensity); ShBi = blue color score for shoulder (adjusted for intensity); LMA = longissimus muscle area obtained from LVS-CCC; PF = overall percent fat in the 12th-/13th-rib interface obtained from VIA output; Sh:Ra ratio = ratio of the maximum rack width and maximum shoulder width; Tw ratio = ratio of the minimum body width (shoulder, rack, loin) and the maximum body width; Lw1 = leg width measurement closest to the groin; and GrAng1 = first groin angle measurement. b Saleable meat yield = subprimal cuts and lean trim from the leg, loin, rack, shoulder, and thin cuts as a percentage of chilled carcass weight. c Subprimal yield = subprimal cuts from the leg, loin, rack, and shoulder as a percentage of carcass side weight. d Fat yield = percentage of chilled side weight of fat from the production of subprimal cuts. View Large Table 7. Descriptive statistics for LM area (cm2) values by method of measurement Method  n  Mean  SD  Minimum  Maximum  CSU—griddeda  147  16.54  2.15  12.10  22.42  LVS-CCC image 1 onlyb  149  16.38  2.33  12.08  22.90  Method  n  Mean  SD  Minimum  Maximum  CSU—griddeda  147  16.54  2.15  12.10  22.42  LVS-CCC image 1 onlyb  149  16.38  2.33  12.08  22.90  a Measurement of the LM area measured by Colorado State University (CSU) personnel using a grid. b Measurement of the LM area measured by lamb vision system-cold carcass component (LVS-CCC), using only picture-1 of each carcass. View Large Table 7. Descriptive statistics for LM area (cm2) values by method of measurement Method  n  Mean  SD  Minimum  Maximum  CSU—griddeda  147  16.54  2.15  12.10  22.42  LVS-CCC image 1 onlyb  149  16.38  2.33  12.08  22.90  Method  n  Mean  SD  Minimum  Maximum  CSU—griddeda  147  16.54  2.15  12.10  22.42  LVS-CCC image 1 onlyb  149  16.38  2.33  12.08  22.90  a Measurement of the LM area measured by Colorado State University (CSU) personnel using a grid. b Measurement of the LM area measured by lamb vision system-cold carcass component (LVS-CCC), using only picture-1 of each carcass. View Large Accuracy of LMA measurements was assessed by regressing measurements from each of the three images obtained using the “triple-trigger” procedure (Steiner et al., 2003) with the LVS-CCC on the Gold Standard measurements of LMA collected by Colorado State University personnel using a grid. Coefficients of determination (R2) and residual standard deviations (RSD) for “Gold Standard” LMA regressed on LVS-CCC measured LMA for each of the three images are presented in Table 8. Prediction equations explained 59, 56, and 60% of the variation in LMA measured by LVS-CCC in images 1, 2, and 3, respectively. The R2 values could have been higher if variability in the “Gold Standards” LMA had been controlled by taking more than one measurement and then averaging them, or if a more accurate method had been used to define “Gold Standard” LMA (e.g., acetate paper tracing and planimeter measurements; Steiner et al., 2003). Table 8. Coefficient of determination (R2) and residual standard deviation (RSD) for “Gold Standards” LM area (LMA, cm2) regressed on lamb vision system-cold carcass component (LVS-CCC) measurements of LMA (cm2)a Item  n  R2  RSD  Image 1 onlyb  149  0.59  1.38  Image 2 onlyc  149  0.56  1.43  Image 3 onlyd  149  0.60  1.36  Item  n  R2  RSD  Image 1 onlyb  149  0.59  1.38  Image 2 onlyc  149  0.56  1.43  Image 3 onlyd  149  0.60  1.36  a LMA measured by Colorado State University personnel using a grid. b Measurement using the first picture of the set of three in the triple-trigger procedure. c Measurement using the second picture of the set of three in the triple-trigger procedure. d Measurement using the third picture of the set of three in the triple-trigger procedure. View Large Table 8. Coefficient of determination (R2) and residual standard deviation (RSD) for “Gold Standards” LM area (LMA, cm2) regressed on lamb vision system-cold carcass component (LVS-CCC) measurements of LMA (cm2)a Item  n  R2  RSD  Image 1 onlyb  149  0.59  1.38  Image 2 onlyc  149  0.56  1.43  Image 3 onlyd  149  0.60  1.36  Item  n  R2  RSD  Image 1 onlyb  149  0.59  1.38  Image 2 onlyc  149  0.56  1.43  Image 3 onlyd  149  0.60  1.36  a LMA measured by Colorado State University personnel using a grid. b Measurement using the first picture of the set of three in the triple-trigger procedure. c Measurement using the second picture of the set of three in the triple-trigger procedure. d Measurement using the third picture of the set of three in the triple-trigger procedure. View Large Repeatability of LMA measurements was assessed by calculating the mean absolute differences for LMA using two methods of LVS-CCC measurement. Mean absolute differences for the “triple-trigger” and “three-placements” methods were 0.13 and 0.15 cm2, respectively. The small mean absolute difference and standard deviation values of 2.35 and 2.41 for the two methods, respectively, suggest that LVS-CCC is highly repeatable in measuring LMA. Variance components and the percentage of total variance in repeated LMA measures accounted for by independent models for “triple-trigger” and “three-placements” methods are presented in Table 9. In both methods of measuring LMA using LVS-CCC, more than 98% of the total variance in LMA was accounted for by the differences between carcasses (repeatability = >0.98). Results of this study suggested that, although there is room for improvement in the accuracy of the measurements by the use of LVS-CCC, the repeatability of the measurements in application was excellent, and this system has the ability to consistently measure LMA. Table 9. Variance components of all LM area (LMA, cm2) measurements using lamb vision system-cold carcass component (LVS-CCC)     Variance due toa  Percentage of total variance due to    Item Method of measurement  n  Carcass ( \({\sigma}^{2}_{c}\) )  Error (σ2ε)  Total  Carcass  Error  Repeatabilityb  Triple triggerc  149  5.51  0.08  5.59  98.6  1.4  0.986  Three placementsd  149  5.77  0.09  5.86  98.4  1.6  0.984      Variance due toa  Percentage of total variance due to    Item Method of measurement  n  Carcass ( \({\sigma}^{2}_{c}\) )  Error (σ2ε)  Total  Carcass  Error  Repeatabilityb  Triple triggerc  149  5.51  0.08  5.59  98.6  1.4  0.986  Three placementsd  149  5.77  0.09  5.86  98.4  1.6  0.984  a \({\sigma}^{2}_{c}\) = variance due to the carcass; σ2ε = residual variance. b Repeatability = carcass variance ( \(s^{2}_{c}\) )/total variance. c Measurement obtained from three sequential images per LM without moving the camera head between pictures. d Measurement obtained from three images per LM removing the camera head between pictures. View Large Table 9. Variance components of all LM area (LMA, cm2) measurements using lamb vision system-cold carcass component (LVS-CCC)     Variance due toa  Percentage of total variance due to    Item Method of measurement  n  Carcass ( \({\sigma}^{2}_{c}\) )  Error (σ2ε)  Total  Carcass  Error  Repeatabilityb  Triple triggerc  149  5.51  0.08  5.59  98.6  1.4  0.986  Three placementsd  149  5.77  0.09  5.86  98.4  1.6  0.984      Variance due toa  Percentage of total variance due to    Item Method of measurement  n  Carcass ( \({\sigma}^{2}_{c}\) )  Error (σ2ε)  Total  Carcass  Error  Repeatabilityb  Triple triggerc  149  5.51  0.08  5.59  98.6  1.4  0.986  Three placementsd  149  5.77  0.09  5.86  98.4  1.6  0.984  a \({\sigma}^{2}_{c}\) = variance due to the carcass; σ2ε = residual variance. b Repeatability = carcass variance ( \(s^{2}_{c}\) )/total variance. c Measurement obtained from three sequential images per LM without moving the camera head between pictures. d Measurement obtained from three images per LM removing the camera head between pictures. View Large Implications Results of the current study confirm previously published information suggesting that the use of an on-line, dual-component lamb vision system can explain a greater proportion of the observed variation in yields of bone-in cuts of lamb vs. expert or on-line USDA yield grades. In addition, results suggest that the lamb vision system-chilled carcass component has the ability to accurately and repeatedly measure longissimus muscle area in ribbed lamb carcasses. Thus, the lamb vision system could be a valuable tool for assisting in the development and implication of a value-based pricing system for sheep in the United States. Literature Cited Brady, A. S., K. E. Belk, S. B. LeValley, N. L. Dalsted, J. A. Scanga, J. D. Tatum, and G. C. Smith 2003. An evaluation of the lamb vision system as a predictor of lamb carcass red meat yield percentage. J. Anim. Sci.  81: 1488– 1498. Google Scholar CrossRef Search ADS PubMed  Montgomery, D. C. 1997. Design and Analysis of Experiments.  4th ed. John Wiley & Sons, New York. Purcell, W. D. 1995. Economic issues and potentials in lamb marketing: Keys to the future of the sheep industry. Sheep Goat Res. J.  11: 92– 105. Steiner, R., D. J. Vote, K. E. Belk, J. A. Scanga, J. W. Wise, J. D. Tatum, and G. C. Smith 2003. Accuracy and repeatability of beef carcasses longissimus muscle area measurements. J. Anim. Sci.  81: 1980– 1988. Google Scholar CrossRef Search ADS PubMed  Steiner, R., A. M. Wyle, D. J. Vote, D. L. Roeber, R. C. Cannell, R. J. Richmond, K. Markey, J. W. Wise, M. E. O'Connor, R. R. Jones, K. E. Belk, J. D. Tatum, and G. C. Smith 2000. Real time augmentation of USDA Yield Grades application to beef carcasses using state-of-the-art VIA instrumentation. Final report to National Cattleman's Beef Association, Englewood, CO.  Colorado State University, Fort Collins. USDA 1992. United States Standards for Grade of Lamb, Yearling Mutton, and Mutton Carcasses. Agric.  Marketing Service, Livestock and Seed Division, USDA Washington, DC. USDA 1996. Institutional Meat Purchase Specifications for Fresh Lamb and Mutton. Agric.  Marketing Service, USDA, Washington, DC. Ward, C. E. 1998. Slaughter lamb pricing issues, evidence and future needs. Sheep Goat Res. J.  14: 35– 42. Copyright 2004 Journal of Animal Science
TI - Development and validation of equations utilizing lamb vision system output to predict lamb carcass fabrication yields
JF - Journal of Animal Science
DO - 10.2527/2004.8272069x
DA - 2004-07-01
UR - https://www.deepdyve.com/lp/oxford-university-press/development-and-validation-of-equations-utilizing-lamb-vision-system-79zLp0nKCK
SP - 2069
EP - 2076
VL - 82
IS - 7
DP - DeepDyve
ER -